Retinoic Acid Mediates Long-Paced Oscillations in Retinoid Receptor Activity: Evidence for a Potential Role for RIP140

NT2/D1, MCF-7, T47D, U2OS and NIH-3T3 cells were purchased from American Type Culture Collection (Manassas, VA). COS-1 cells were a gift from Dr. John Hwa (Dartmouth Medical School). MCF-7 and T47D cells were cultured in DMEM/F12 (50:50) and all other cell lines were cultured in DMEM (Gibco).

For estrogen treatments, MCF-7 cells were cultured in phenol-free medium beginning 24 hours before drug treatment. 17-β−estradiol (E2) and RA were purchased from Sigma. 1,25-dihydroxyvitamin D₃ was purchased from Biomol. Dexamethasone and rosiglitazone were gifts from Dr. Joshua Hamilton (Dartmouth Medical School) and Dr. Michael Sporn (Dartmouth Medical School), respectively. RA was stored protected from light in liquid nitrogen as a 10 mM stock solution in dimethyl sulfoxide (DMSO). 4-[(E)-2-(5,6,7,8-Tetrahydro-5,5,8,8-tetramethyl-2-naphthalenyl)-1-propenyl] benzoic acid (TTNPB, Tocris Bioscience) and bexarotene (LC Laboratories) were stored away from light at −20°C as 10 mM stock solutions in DMSO. All other drugs were dissolved in ethanol (EtOH) and stored away from light at −20°C. For the extended time courses of RA treatment, 1×10⁶ NT2/D1 cells were plated in 10 cm plates and maintained in charcoal-absorbed sera containing media for 48 hours to deplete endogenous NR ligands. After 48 hours, media was replaced with serum-free media containing 2.5 µM α-amanitin (Sigma) for 2 hours. Cells were washed and DMEM containing 10% charcoal-absorbed sera and 1 µM RA was added. RA remained on the cells until harvest.

For siRNA experiments, custom designed siRNA duplexes for RIP140, NCoR1 and SRC1 were purchased (Dharmacon, Lafayette, CO). Human RIP140 siRNA is of sequence 5′-CAAACAGGAUAGCACAUUA-3′ and corresponds to the RIP140 cDNA beginning 1986 bp downstream of the ATG start codon. Human NCoR1 siRNA is of sequence5′-AAAGUCGUUAUCCUCCUCACU-3′ and corresponds to the NCoR1 cDNA beginning 49 bp downstream of the ATG start codon. Human SRC1 siRNA is of sequence 5′-AGGAGACAGGUUACUUCUG-3′ and corresponds to the cDNA beginning 1467 bp downstream of the start codon. The control was the Scrambled siRNA of sequence 5′-GCGCGCUUUGUAGGAUUCG-3′ from Dharmacon. For transient siRNA experiments, log-phase cells were transfected with 150 nM siRNA using OligofectAMINE (Invitrogen) as described [11], [12]. Briefly, 2−3×10⁶ cells per 10 cm plate were transfected with 150 nM siRNA for 16 h. Cells were split the following day onto 10 cm plates with charcoal-absorbed sera media containing either 1 µM RA or DMSO. Cells remained in this media until the time of harvest.

The pSG5-RIP140 expression plasmid has been described previously [10], [11]. The pCMX-PPARγ expression plasmid and the reporter constructs PPRE-tk-Luc and VDRE-tk-Luc were kindly provided by Dr. Michael Sporn (Dartmouth Medical School). The Ebo-RXR expression plasmid was a gift from Dr. Ethan Dmitrovsky (Dartmouth Medical School). The pCI-mGR expression plasmid contains the full-length murine glucocorticoid receptor sequence and was kindly provided by Dr. Lynn Sheldon (Dartmouth Medical School). For reporter assays, COS-1, NT2/D1, or T47D human breast cancer cells were plated at 0.2×10⁶ cells per well of a 6-well plate and transfected with 2 µg total DNA using Polyfect™ (Qiagen) according to the manufacturer's instructions. Transfections were performed in triplicate for each condition.

For the PPRE-Luc experiment, COS-1 cells were transfected with 0.1 µg PPRE-Luc, 0.125 µg CMV-PPARγ, 0.125 µg Ebo-RXRA, 0.15 µg pSG5-RIP140, and 0.1 µg Renilla luciferase plasmid (pRL-tk) per well. For the GRE-luc experiment, COS-1 cells were transfected with 0.1 µg GRE-Luc, 0.1 µg GR, 0.1 µg pRL-tk, and 0.45 µg pSG5-RIP140. The dual-luciferase assay system (Promega) was used to normalize firefly luciferase activity against renilla luciferase activity. In all conventional reporter transfections, total DNA was brought to 2 µg per well with salmon sperm DNA. Media containing the transfection mixture was removed after 16 hours and charcoal-absorbed sera media containing 1 µM rosiglitazone, 1 µM dexamethasone, 1 µM vitamin D₃ or vehicle control (DMSO or ethanol) was added. Cells were lysed 24 hours following drug treatment and luciferase activity was determined. Data points represent the average of triplicate transfections.

The RARE-tk-Luc reporter consists of a fusion of two RARE sequences to a minimal herpes simplex virus-thymidine kinase promoter and luciferase and has been used previously [10], [11]. For transient transfection real-time reporter assays, 1.5×10⁴ NT2/D1 cells were plated in triplicate 35 mm plates in charcoal absorbed sera media 24 hours prior to transfection. Cells were transfected with 1.75 µg total DNA per plate consisting of 0.75 µg RARE-tk-Luc and 1 µg insertless pSG5 or pSG5-RIP140. For the siRNA condition, 150 nM RIP140 siRNA was transfected with 0.75 µg RARE-tk-Luc with otherwise identical transfection parameters as the other three conditions. Transfections were performed using Polyfect (Qiagen) according to the manufacturer's directions. After a 16 hour incubation with the transfection mixture, cells were washed with PBS and 2 ml of Leibovitz's L15 media (Gibco) containing 10% charcoal-absorbed FBS, 1% penicillin/streptomycin, 10 mM HEPES, 0.1 µM luciferin and 1 µM RA or DMSO was added to each plate. Plates were sealed with silicone grease and glass cover slips and immediately transferred to a LumiCycle (Actimetrics, Wilmette, IL) for data collection as previously described [14].

For real-time reporter assays employing U2OS and NIH-3T3 cells containing a stably integrated RARE-tk-luciferase cassette, cells were plated to confluency in triplicate 35 mm plates in charcoal absorbed sera media. The following day, cells were transferred to 0.5% charcoal absorbed sera media for 2 days. For one arm of the U2OS experiment cells were pretreated with 2.5 µM α-amanitin in serum-free media for 2 hours. All cells were then washed with PBS and 2 ml of Leibovitz's L15 media containing 10% charcoal-absorbed FBS, 1% penicillin/streptomycin, 10 mM HEPES, 0.1 µM luciferin and 10 µM RA or 5 µM TTNPB plus 5 µM bexarotene was added to each plate. Plates were sealed and transferred to the LumiCycle for data collection. The data collection interval was set for 6 collections/hour. The LumiCycle is housed in a standard tissue culture incubator at 36°C instead of 37°C to minimize background counts. The data was collected up to 72 to 100 hours and analyzed with the LumiCycle Analysis software (Actimetrics). For the U2OS and NIH-3T3 cells, the data were baseline subtracted based on a running average as previously described [14].

The stable RARE-tk-Luc U2OS and NIH3T3 cells were generated using a lentivirus expressing luciferase under control of the above described RARE-tk promoter and enhancer. The lentiviral RARE-tk-Luc plasmid was obtained by cloning the RARE-tk portion of the RARE-tk-Luc plasmid in place of the β-catenin response element in the Topflash lentiviral vector obtained from Dr. Karl Willert (University of California). DNA sequencing confirmed successful cloning.

For generation of stable RIP140 knockdown in NT2/D1 cells, a RIP140 shRNA lentiviral construct with a sense strand sequence of 5′-GCGGAGAAGAATGAGTATGAA-3′ was purchased (Open Biosystems, clone ID TRCN0000019779). A pLKO.1-shRNA lentiviral construct targeting eGFP of sense strand sequence 5′-TACAACAGCCACAACGTCTAT-3′ was also purchased (pLKO.1-eGFP, Sigma) and used as a control. Lentiviral stocks were generated from 293T cells and psPAX2 and pMD2G packaging and envelop vectors using standard protocols. Viral stocks were added to NT2/D1 cells plated at 0.2×10⁶ cells/well per 6 well plates. The lentiviral stock remained on the cells for 24 h. Cells were split onto 10 cm plates in selection media containing 0.5 µg/ml puromycin. Selection continued for 48 hours, after which no viable cells remained in a mock-transduced control plate. The puromycin-resistant cell pools were used for experiments.

Total RNA was isolated using TriReagent (Molecular Research Center). Northern hybridizations were performed with 5 µg RNA as described previously [10], [11] using radiolabeled 500–1100 bp DNA fragments from RIP140, SRC1, NCoR1 or RARB. The RIP140 probe was a 1000 bp fragment from a BglII/XhoI digest of the pSG5-RIP140 plasmid [10], [11]. The SRC1 probe was a 1071 bp fragment from a BamHI/HindIII digest of the pCR3.1-hSRC1A construct kindly provided by Dr. Bert O' Malley (Baylor College of Medicine). The NCoR1 probe was a 1075 bp fragment of the pCMX-NCoR construct provided by Dr. Ronald Evans (The Salk Institute). For semiquantitative PCR, cDNA was synthesized from 5 µg RNA using Superscript II reverse transcriptase (Invitrogen), and CYP24A1 or RIP140 expression was analyzed by PCR using Taq polymerase (Invitrogen) to cycle numbers within the linear range as described [12]. For realtime RT-PCR, cDNA was synthesized from 2 µg RNA using TaqMan reverse transcription reagents (Applied Biosystems) and RIP140, RARA, RARB, SRC1, NCoR1, and HOXA5 were amplified from 25 ng cDNA per reaction using SYBR Green reagent (Applied Biosystems). RIP140 transcripts containing exon 1a and exon 1b were amplified using TaqMan primers and reagents (Applied Biosystems). Gene expression data was normalized to GAPDH. Primer sequences are listed in Figure S1. Data was collected and analyzed with 7500 Fast System Sequence Detection Software version 1.4 (Applied Biosystems).

The Evolutionarily Conserved Regions Browser (http://ecrbrowser.dcode.org↗) is a graphical interface described by Loots et al. [15]. This resource was used to search the RIP140 gene for consensus transcription factor response elements that exhibit a high level of evolutionary conservation. A 550 bp region of the RIP140 ECR was amplified by genomic PCR and cloned into a TK-luciferase construct. PCR primers were 5′ TAGGTAAACCCTCTCTCCAGAAT and 5′ ATCCTCTCCATCAACTACAAAGC. DNA sequencing confirmed that the correct sequence was cloned. The cloned sequence can be found under the previously existing NCBI accession number, NT_011512↗.

We sought to characterize RAR activity over an extended time course in the constant presence of RA. To this end, NT2/D1 cells were transiently transfected with a RA-responsive luciferase reporter (RARE-tk-Luc), and RA-mediated RARE transactivation was monitored in real time with a LumiCycle instrument. RA-dependent reporter activity was monitored in biological triplicate every 10 minutes for 76 hours following addition of RA. RA treatment resulted in oscillatory RAR activity with an extended period (Figure 1A and 1B). We believe this is the first demonstration of long-timed oscillations in RAR activity in the continued presence of RA.

Our previous studies using conventional reporter assays established that RIP140 siRNA enhances RAR activation of endogenous RA target genes and heterologous RARE-containing promoters while RIP140 overexpression represses this activation [10]–[12]. These studies involved only one time point (generally 24 or 48 hours). We sought to characterize whether the oscillations in RAR activity were influenced by RIP140 by employing RIP140 siRNA or a RIP140 expression plasmid (HA-RIP140). RARE–dependent oscillations in the constant presence of RA were completely abolished by RIP140 overexpression driven by a constitutive SV40 promoter that is not subject to RA regulation (Figure 1A). The linear decline in RAR activity may reflect accumulation of RIP140 over the course of the experiment.

RIP140 siRNA produced an overall amplification of RARE-luciferase activity as compared to RA treatment alone. RAR activity was robustly enhanced by RIP140 siRNA at all time points beginning at 4 hours (Figure 1A). However, the oscillations were slightly perturbed with a suggestion of a partially “flattened” response (Figure 1B). Our previous work indicated that RIP140 siRNA does not completely ablate basal or RA-induced RIP140 expression in NT2/D1 cells [11], [12]. This residual RIP140 expression may account for the retention of RAR oscillation with siRNA treatment.

Based on these data the following hypothetical model depicted in Figure 1C is proposed. We have previously found that RA induction of RIP140 constitutes a negative feedback mechanism that limits and fine-tunes retinoid signaling [10]–[12]. The negative feedback predicts that upon RA addition, RIP140 should accumulate until a level sufficient to inhibit RAR activity is attained. RIP140 expression should then decline, allowing a recovery of RAR transactivation activity and the re-expression of RIP140.

A relationship between the time required to reach the initial peak of RARE activity and RIP140 status was also evident. With RIP140 overexpression the peak was reached most rapidly, at 5 hours (“c” in Figure 1A), whereas with endogenous or intermediate RIP140 levels, the peak was reached at 7 hours (“b” in Figure 1A). With RIP140 depletion, the peak was reached at 13 hours (“a” in Figure 1A).

To further examine RAR oscillatory activity, U2OS cells containing a stably integrated RARE-tk-luciferase cassette were utilized (Figure S2). In contrast to the NT2/D1 cells which have decreased viability upon prolonged confluence, U2OS cells were allowed to grow to confluence and serum-starved for 2 days prior to ligand addition. This allowed for the assessment of time-dependent nuclear receptor activity independent of the cell cycle. In addition to RA, U2OS cells were also treated with a combination of the retinoid analogs, TTNPB and bexarotene, which are RAR and RXR selective agonists, respectively. These ligands were chosen because they are reported to be more stable than RA and because RIP140 has been reported to more strongly bind to the RAR/RXR heterodimer when both RAR and RXR are engaged with ligand [16]. In addition, a group of cells were first treated with 2.5 µM α-amanitin for 2 hours prior to TTNPB and bexarotene treatment in an attempt to “transcriptionally synchronize” the cells [17]–[19]. For all treatment protocols, an initial peak in RARE-luciferase activity was evident after ∼10 hours of retinoid treatment followed by a second peak (35–45 hours) and a third peak (∼70 hours), resulting in a period ranging from 25–35 hours (Figure S2A). There was little difference between the RA and bexarotene/TTNPB treatments, and α-amanitin pretreatment did not affect the results. The amplitude of the oscillations dampened over the duration of retinoid treatment (Figure S2A). Dampening is often seen in biological oscillators and depends on the relative values of feedback parameters and also results from progressive asynchronization and averaging over a cell population [20]. No oscillations were seen when cells were treated with the vehicle control DMSO in place of RA (data not shown). These data indicate that RAR oscillatory activity from an integrated reporter can be maintained in the constant presence of RA even in cells that have left the cell cycle and are no longer proliferating.

We next examined the dynamics of RAR activation in a third cell context, NIH-3T3 cells stably integrated with RARE-tk-luciferase. NIH-3T3 cells were confluent and serum starved and therefore not proliferating during the course of bexarotene and TTNPB treatment. Luciferase activity first peaked between 8 and 12 hours, followed by two subsequent peaks at 26–32 and 60–65 hours for a period of ranging from 25–30 hours (). As seen above in U2OS cells, the amplitude of the oscillations dampened over the duration of the retinoid treatment. Figure S2B

The negative feedback predicts that RIP140 should cycle in the constant presence of RA. To investigate whether RIP140 mRNA cycles with continual RA exposure, we pretreated NT2/D1 cells with the RNA polymerase II inhibitor, alpha-amanatin, for 2 hours and then treated the cells with RA over an extensive 57-hour time course. Treatment with alpha-amanatin is a common strategy to synchronize NR activity and has been used extensively to measure rapid ligand-dependent coactivator complex recruitment to promoters of NR target genes [17]–[19]. Steady-state RIP140 mRNA increased over the first 25 hours of RA treatment, followed by a steady decline over the next 10 hours and a partial recovery from 40 to 57 hours of RA treatment (Figure 2A). Primary unspliced transcript levels also cycled with increased amplitude (Figure 2B). The oscillations were not observed in cells treated with alpha-amanatin alone. These oscillations are consistent with the negative feedback model and long 24–30 hour periodicity of RIP140 expression is consistent with long-timed transcriptional-translational feedback. Further it is possible that the oscillation is circadian in nature since the period partially overlaps with a circadian period and given the known relationship between NRs and the circadian clock apparatus [21]. Whether RIP140 would eventually decline over time due to negative feedback of RAR activity would require longer time points.

Cyclic RAR activation suggests that endogenous RA target genes should also oscillate with continued RA exposure. As depicted in Figure 3, both primary and steady state RARB transcripts oscillate with a period of approximately 25–30 hours in the continued presence of RA. HOXA5 expression was also cyclic with a similar period as RARB (Figure 3). For both RARB and HOXA5, oscillations of primary transcript levels, which are more indicative of transcription rate, were more pronounced as compared to steady state mRNA levels.

RIP140 constitutes a negative feedback mechanism to limit RA signaling and is expressed in a cyclic fashion with continued RA treatment. We asked whether RIP140 depletion perturbs the oscillation of an RA target gene. The effect of RIP140 siRNA on the cyclic expression of HOXA5 was explored. NT2/D1 cells were transiently transfected with control or RIP140 siRNA. Cells were subjected to a 56-hour time course of RA treatment. RNA was harvested every 4 hours and gene expression was determined by real-time PCR. Under these conditions RA induced a clear oscillation in HOXA5 expression with a period approximating 30 hours in control siRNA treated cells (Figure 4). RIP140 siRNA potentiated HOXA5 induction after RA exposure. The magnitude of HOXA5 expression by RIP140 siRNA increased dramatically as compared to control siRNA treated cells. At 56 hours, HOXA5 expression increased ∼20-fold with RA in control siRNA cells and ∼140-fold with RA in RIP140 siRNA cells (Figure 4A). Consistent with data in Figure 3D, HOXA5 expression oscillated in control cells (Figure 4B). The kinetics of HOXA5 expression in RIP140 siRNA cells were altered such that HOXA5 oscillations were replaced by a continual increase in HOXA5 expression over time. This induction had two phases, a modestly increased rate of HOXA5 expression as compared to control cells within in the first 35 hours followed by a steep increase in the expression of HOXA5 thereafter (Figure 4A). These data suggest a potential role for RIP140 in fine-tuning the oscillatory expression of HOXA5 in the continued presence of RA.

Our prior work established that RIP140 siRNA enhances and accelerates retinoid receptor transactivation, expression of RA target genes, and RA-mediated differentiation and growth suppression of pluripotent human embryonal carcinoma cells [11], [12]. Interestingly, microarray studies suggested that distinct subsets of direct RA target genes differentially responded to RIP140 depletion [12]. We wished to validate this apparent selectivity of RIP140.

NT2/D1 cells were transduced in triplicate with a lentiviral short hairpin RNA (shRNA) targeting RIP140, pLKO.1-RIP140, or a control shRNA targeting GFP, pLKO.1-shGFP. RIP140 shRNA mediated substantial RIP140 depletion under both basal and RA-treated conditions and resulted in a substantial enhancement of RA dependent expression of RARB (Figure 5A). One shRNA pool and one control pool were used to evaluate the kinetics of RARA and RARB induction with RA. Real-time PCR indicated enhancement of RA-dependent RARB expression but no enhancement of RARA expression in RIP140 shRNA cells as compared to control shRNA cells (Figure 5B–5D).

Similarly, RIP140 knockdown using siRNA enhanced RA induction of RARB and HOXA5 but not RARA in NT2/D1 cells (Figure S3 and data not shown). Further, both shRNA and siRNA knockdown of RIP140 enhanced RA induction of RARB but not RARA in the RA-responsive murine lung adenocarcinoma line, ED-1 [22] (data not shown). Thus, in two different cell systems with both siRNA and shRNA knockdown, RIP140 limits RARB and HOXA5 but not RARA expression despite the fact that all three genes are well-characterized, RARE-containing, direct targets of RARs.

We wished to begin to address whether RIP140 plays a distinct role in the regulation of RA signaling as compared to classic coregulators of nuclear receptors, the p160 family of coactivators (SRC1, SRC2 and SRC3) and the classic corepressors (NCoR1 and NCoR2/SMRT). In contrast to RIP140 siRNA, siRNA to NCoR1 and SRC1 failed to affect RA-dependent RARB induction in NT2/D1 cells (Figure 6A and 6B). In addition, knocking down RIP140, SRC1 or NCoR1 did not greatly influence the expression of one another (Figure 6A, 6C and 6D). While RIP140 is RA inducible and cycles in the constant presence of RA (Figure 2), NCoR1 and SRC1 expression is not responsive to RA and does not oscillate in the constant presence of RA (Figure 6 and Figure S4). These data suggest that RIP140 plays a distinct and non-redundant role in RA-signaling in NT2/D1 cells as compared to SRC1 and NCoR1, which may share roles with the additional known homologs SRC2, SRC3 and NCoR2.

We have shown previously that a four-fold induction of RIP140 by RA occurs within 3 hours of RA treatment and de novo protein synthesis is not required for this induction, indicating that RIP140 is a RA target gene [10]. Gene expression profiling has identified the upregulation of RIP140 by RA, estrogens, and vitamin D₃[23]–[25]. Estrogens and androgens directly induce RIP140 in MCF-7 breast cancer cells and LNCaP prostate cancer cells, respectively [26], [27]. To investigate the inducibility of RIP140 by nuclear receptor ligands, MCF-7 cells were treated with RA, 17β-estradiol, dexamethasone, vitamin D₃, rosiglitazone or vehicle control for 24 hours and expression of RIP140 was analyzed by real-time RT-PCR. MCF-7 cells were chosen since they are known to express the cognate receptors for these ligands. RA caused the greatest accumulation of RIP140 mRNA while estradiol and rosiglitazone caused a more modest induction (Figure 7B). A slight induction of RIP140 was also seen with vitamin D₃ in MCF-7 cells (Figure 7B).

The RIP140 locus spans approximately 100 kb and consists of one coding exon (exon 5) and several upstream non-coding exons (Figure 7A). Multiple RIP140 promoters have been identified and alternative splicing yields multiple RIP140 transcripts [25], [26], [28]. In MCF-7 cells, transcripts containing 1b (Figure 7C) and 1a (data not shown) accumulated after 24 hours of treatment with RA, estrogen or rosiglitazone. Utilizing the Evolutionarily Conserved Regions (ECR) browser [15], we located a 3.5 kb region between exons 3 and 4a that is the most highly evolutionarily conserved region in the RIP140 locus aside from the protein coding sequence contained within exon 5 (Figure 7A). The TRANSFAC algorithm indicated that this region contains sequences matching half or full consensus site response elements for several NRs including ERR, FXR, LXR, ROR, GR and also a consensus ERa/Sp1 type ER responsive element.

We cloned a 550 bp fragment of this region into a thymidine kinase (tk)-luciferase construct and cotransfected this plasmid along with an ERα expression plasmid in COS-1 cells. Estrogen treatment for 24 hours resulted in an approximate 8-fold induction in reporter activity, supporting the presence of a functional ERE in this region (Figure 7D). However, the construct was not responsive to RA or dexamethasone, indicating that functional GREs and RAREs do not exist in this region of the RIP140 locus (data not shown).

There is growing evidence to support a role for RIP140 as a global ligand-dependent NR corepressor [29]. We sought to determine the effect of exogenous RIP140 on the activity of heterologous GR- and PPARγ-responsive promoters (Figure 8). RIP140 cotransfected with GR or PPARγ in COS-1 cells inhibited GRE- and PPRE-dependent activation by dexamethasone and rosiglitazone, respectively (Figure 8A and 8B). In addition, induction of CYP24A1, an endogenous vitamin D₃ target gene with a well-defined VDRE, was enhanced with RIP140 siRNA in T47D cells (Figure 8C). These results, along with the data on RIP140 inducibility by NRs (Figure 7) supports prior studies suggesting that RIP140 may be a general ligand-dependent and ligand-inducible repressor of NRs [6], [23]–[31]. These properties suggest that RIP140 has the potential to be involved in oscillatory transactivation of other NRs in addition to RARs.