Disruption of Retinoic Acid Receptor Alpha Reveals the Growth Promoter Face of Retinoic Acid

Background Retinoic acid (RA), the bioactive derivative of Vitamin A, by epigenetically controlling transcription through the RA-receptors (RARs), exerts a potent antiproliferative effect on human cells. However, a number of studies show that RA can also promote cell survival and growth. In the course of one of our studies we observed that disruption of RA-receptor alpha, RARα, abrogates the RA-mediated growth-inhibitory effects and unmasks the growth-promoting face of RA (Ren et al., Mol. Cell. Biol., 2005, 25:10591). The objective of this study was to investigate whether RA can differentially govern cell growth, in the presence and absence of RARα, through differential regulation of the “rheostat” comprising ceramide (CER), the sphingolipid with growth-inhibitory activity, and sphingosine-1-phosphate (S1P), the sphingolipid with prosurvival activity. Methodology/Principal Findings We found that functional inhibition of endogenous RARα in breast cancer cells by using either RARα specific antagonists or a dominant negative RARα mutant hampers on one hand the RA-induced upregulation of neutral sphingomyelinase (nSMase)-mediated CER synthesis, and on the other hand the RA-induced downregulation of sphingosine kinase 1, SK1, pivotal for S1P synthesis. In association with RA inability to regulate the sphingolipid rheostat, cells not only survive, but also grow more in response to RA both in vitro and in vivo. By combining genetic, pharmacological and biochemical approaches, we mechanistically demonstrated that RA-induced growth is, at least in part, due to non-RAR-mediated activation of the SK1-S1P signaling. Conclusions/Significance In the presence of functional RARα, RA inhibits cell growth by concertedly, and inversely, modulating the CER and S1P synthetic pathways. In the absence of a functional RARα, RA–in a non-RAR-mediated fashion–promotes cell growth by activating the prosurvival S1P signaling. These two distinct, yet integrated processes apparently concur to the growth-promoter effects of RA.


INTRODUCTION
Retinoic acid (RA), the bioactive derivative of dietary Vitamin A and beta-carotene, is a powerful signaling molecule controlling cell proliferation [1]. Normally, RA exerts an inhibitory action on cell proliferation via specialized transcription factors, the nuclear RA receptors, RARs [2]. Cells evolved an amazing apparatus to safely control the RA antiproliferative action. First, cells finely regulate the level of intracellular RA through a sophisticated metabolic/ homeostatic process [3]; second, they use specialized cellular retinoic acid binding proteins (CRABPs) to chaperone RA from the cytoplasm directly onto the RARs in the nucleus [4]; third, they have evolved a two-tier RAR-regulated system to control the downstream transcription of genes in response to RA [5]. RA, after binding the nuclear receptor RARa, triggers the transcription of other downstream RARs, including the RA-receptor and tumor suppressor, RARb2 [6]. RARb2 autoregulates its own transcription and, in turn, the transcription of a multitude of downstream RAresponsive target genes [7,8].
According to several literature reports, RA and its dietary precursors can also promote, rather than inhibit, cell survival and growth [18][19][20][21][22]. In the course of a recent study, we observed that disruption of RARa signaling in RA-sensitive breast cells not only leads to RA-resistance, but unexpectedly unmasks the growthpromoter face of RA [13,[23][24][25].
Here we show that in RA-sensitive cells with a functional RARa signaling RA leads to growth inhibition consequent to the concerted upregulation of neutral sphingomyelinase (nSMase), one of the enzymes leading to the synthesis of the antiproliferative/propaptotic ceramide (CER), and downregulation of sphingosine kinase 1 (SK1), the enzyme leading to the synthesis of the prosurvival sphingosine-1-phosphate (S1P). In contrast, disruption of RARa signaling in the same cells results, in response to RA, into increased proliferation associated with both loss of concerted regulation of nSMase and SK1, and induction of intracellular S1P. Altogether our findings indicate that the presence of RARa is essential for the proper regulation of the sphingolipid rheostat by RA. In the absence of RARa, RA no longer executes its growth-inhibitory action through its canonical receptors, but activates the prosurvival SK1-S1P pathway through alternate non-RAR receptor(s).

RESULTS
Cells with a functionally disrupted RARa signaling become both RA-resistant and susceptible to RA-induced cell growth in vitro and in vivo By using different strategies to functionally inhibit RA-RARa signaling in RA-sensitive breast cancer cells (T47D), we found that-concomitant with heritable epigenetic gene silencing of the downstream RARb2 receptor and tumor suppressor-cells not only survive, but also proliferate significantly more in response to RA ( [13] and Fig. 1A). Specifically, we observed that T47D-derived clones, obtained by stably inhibiting the endogenous RARa signaling with either the RARa-specific antagonist ER50891 (here shown a prototypic clone, ER-C4) (Fig. 1B, top) or the dominant negative DN RARa403 mutant (here shown a prototypic clone, DNC8) (Fig.1B, bottom), were not only RA-resistant, but grew significantly more in the presence of RA (1 mM, 72 h) as shown by both colony formation assay and cell proliferation assay, while their respective controls, T47D in the case of ER-C4, and LXC5 in the case of DNC8, were growth-inhibited by RA. We will present hereafter only the results concerning the LXC5/DNC8 isogenic model, because the T47D/ER-C4 show an identical RAresponse for the different parameters that we analyzed in this study.
The increased RA-induced growth is supported by the observation that cells in the presence of RA (1 mM, 72 h) transition more rapidly from the G1 to the S phase (Fig. 1C, left) in agreement with an increased transcription of the cyclin D1 gene, encoding a protein pivotal for the G1-S phase transition (Fig.1C, right).
RA, and its dietary precursor retinol, can apparently promote DNC8 cell growth in vivo. DNC8 cells xenografted subcutaneously, and bilaterally, in the dorsal flank of female nude mice (see experimental scheme, Fig. 2A), were clearly growth-promoted by chronic RA treatment (2.5 mg/kg) delivered by daily intraperitoneal injection. Weekly assessment of tumor size showed that RA clearly promoted the growth of DNC8 xenograft tumors up to the sixth week (Fig. 2B, right); thereafter tumors stop growing (data not shown). In contrast, the same RA treatment induced growth inhibition of the control LXC5 xenograft tumors (Fig. 2B, left). Immunocytochemistry of DNC8 tumor sections after six-week RA-treatment showed a significantly (p,0.05) higher number of cells positive for the proliferation marker Ki67 (Fig. 2C, left). Significantly (p,0.01) higher was also the level of cyclin D1 transcription (Fig. 2C, right).
Based on the overall in vitro and in vivo observations, we hypothesized that two distinct effects occur as a consequence of disruption of RARa function. The first effect is the abrogation of growth inhibition mediated by RA through RARa, and the second effect is a non-RARa-mediated stimulation of cell growth by RA itself. To identify these effects we focused our analysis on CER and S1P signaling, two sphingolipid signaling pathways exerting opposite action on cell growth.

RA fails to induce CER synthesis in cells with functional RARa inhibition
The metabolism of both CER and S1P is tightly integrated (Fig. 3A). CER can be generated either as a result of sphingomyelin hydrolysis, catalyzed by either one of two sphingomyelinases, the neutral sphingomyelinase (nSMase) and the acid sphingomyelinase (aSMase), or by de novo synthesis as a result of condensation of L-serine and palmitoyl CoA catalyzed by serine palmitoyltransferase (SPT) (Fig. 3A). In preliminary cell labeling experiments of T47D cells with either [ 3 H] sphingosine or [ 3 H] palmitate, it was apparent that RA induces CER synthesis via sphingomyelin hydrolysis, and not de novo synthesis in cells with a functional RARa (Fig. 3B). Consistently, both the transcription level of the two SPT subunits genes, LCB1 and LCB2, and the SPT activity did not vary significantly between LXC5 and DNC8 cells in response to RA (Fig. 3C, top). In contrast, both the transcription level and activity of one of the sphingomyelinases, nSMase (Fig. 3C middle), significantly (p,0.01) increased in response to RA in LXC5, but not in DNC8 cells. Conversely, the transcription and activity of aSMase remained unchanged (Fig. 3C bottom). Moreover, a specific nSMase inhibitor, GW4869 [26] (5 mM, 48 h) significantly (p,0.05) counteracted RA-induced growth inhibition (Fig. 3D, left) as well as RA-induced CER level in LXC5 cells (Fig. 3D, right).
To validate independently whether nSMase-driven CER synthesis is under RARa regulation, we used two specific RARa antagonists RO415253 [27] and ER50891 [28]. Both antagonists effectively inhibit RA action at RARa, since they abrogate RAinduced transcriptional upregulation of RARb2, a prototypic direct RARa-target (Fig. 4C). Treatment of T47D cells with a 100fold concentration of either one of the RARa antagonists relative to RA for 72 h, counteracted both the RA-induced antiproliferative activity (Fig. 4A) and the RA-induced CER synthesis in T47D cells (Fig. 4B). Further, both RARa antagonists inhibited the RAinduced transcriptional upregulation of nSMase (Fig. 4C). These findings demonstrate a functional interference of both antagonists with RA-induced, nSMase-mediated CER synthesis.
Thus, by using two independent approaches, we clearly demonstrated that disruption of RARa function abrogates the nSMase-mediated synthesis of CER in response to RA.
Fenretinide, a retinoid that works in an RAR-independent fashion, can induce CER in cells with functional RARa inhibition Fenretinide (4-HPR) is a synthetic retinoid that was shown to be effective for prevention and treatment of breast cancer [29,30]. Fenretinide was reported to induce CER accumulation in a non-RAR-dependent fashion [31,32]. Consistently, DNC8 cells, while unable of nSMase-induced CER synthesis in response to RA (1 mM, 72 h) (Fig. 5A left), were capable of accumulating CER in response to fenretinide (4 mM, 72h) (Fig. 5A, right). Fenretinideinduced CER accumulation is paralleled both by a consistent antiproliferative (Fig. 5B, left) and proapoptotic effect (Fig. 5B, right).
These observations support the conclusion that the inability of RA to induce CER synthesis in cells with functional disruption of RARa is due to lack of upregulation of the specific nSMasemediated CER synthetic pathway, and not to an overall failure of the entire CER synthetic apparatus.

RA fails to downregulate both SK1 transcription and activity in cells with functional RARa inhibition
The metabolism of the antiproliferative CER is intrinsically linked to the metabolism of the prosurvival S1P effector (Fig. 3A). For this reason, we measured the SK activity in LXC5 and DNC8 cells both at baseline (in the absence of RA) and in the presence of RA. Apparently, LXC5 cells have a significantly (p,0.01) lower level of SK1 activity than DNC8 cells already at baseline (Fig. 6A). Further, RA (1 mM, 72 h) can induce a significant downregulation (p,0.05) of SK1 activity in LXC5 but not in DNC8 cells (Fig. 6A, upper inserts). The SK1 activity pattern in LXC5 and DNC8, at baseline and after RA-treatment, mirrors the transcription pattern of the SK1 gene, one of the two SK genes (Fig. 6B, top left). The transcription of sphingosine kinase 2 (SK2), sphingosine-1-phosphate lyase (S1P lyase), and sphingosine-1-phosphate phosphatase (S1PP) is not significantly different in LXC5 and DNC8 cells both at baseline and after RAtreatment ( Fig. 6B, top right, and bottom).
Apparently, only in cells with a functional RARa, RA transcriptionally regulates in an opposite fashion the metabolic pathways leading to either CER or S1P synthesis, by upregulating on one hand nSMase transcription and by downregulating on the other hand SK1 transcription, thus synergistically inhibiting cell proliferation.
Evidence that RA fails to regulate in an opposite fashion CER synthesis and SK activity in cells lacking endogenous RARa Next, we searched for evidence that RA fails to regulate in an opposite fashion CER synthesis and SK activity also in cells that lack endogenous RARa function. For this reason, we chose a breast cancer cell line, MDA-MB-231, that does not express endogenous RARa (Fig 7A, left) and the other downstream RA-regulated RAR genes, including RARb2 (Fig. 7A, right). MDA-MB-231 are modestly, yet significantly (p,0.05) growth-promoted by RA (Fig. 7B). Interestingly, in these cells RA fails to: induce nSMase Treatment with the SK inhibitor 2-(p-hydroxyanilino)-4-(pchlorophenyl) thiazole (2 mM, 72 h) led to significant inhibition (p,0.01) of MDA-MB-231 proliferation both in the absence and presence of RA (Fig. 7D), indicating that RA-promoted cell proliferation of MDA-MB-231 cells might be due, at least in part, to activation of the SK1-S1P signaling pathway. In addition, MDA-MB-231 can accumulate CER (Fig. 7E, left) and undergo apoptosis (Fig. 7E, right) in response to fenretinide, showing that other non-RAR-regulated ceramide synthetic pathways are still functional. Thus, in different cell contexts, both when we disrupted functional RARa (DNC8) or there is no endogenous RARa (MDA-MB-231) it is apparent that the CER/S1P rheostat is not regulated by RA as it does in cells with an intact RARa. In contrast, RA seems to activate, rather than downregulate, the SK signaling.
SK1-S1P signaling: a candidate growth promoting mechanism of non-RAR-mediated RA action RA-induced cell survival and growth have been documented in different cells and tissues [33][34][35]. RA-induced proliferation in the absence of a functional RARa signaling is not breast cancer cell context-specific and can occur both in transformed and untransformed cells (unpublished observations). Apparently, a few non-RAR targets can mediate RA-action [34,36,37]. We gathered preliminary evidence that in DNC8 cells the RA non-RAR- mediated proliferation effect is due, at least in part, to activation of the SK1-S1P signaling pathway because of the following observations. First, in the presence of RA, cyclin D1 transcription is upregulated in DNC8 cells transfected with wild-type SK1 (Fig. 8A, left). Second, exogenous expression of a dominant negative SK1 mutant (DNSK) in DNC8 cells significantly (p,0.05) reduced the level of cyclin D1 transcription compared to the level of cyclin D1 transcription in cells transfected with the cognate empty vector (Fig. 8A, right). Third, treatment with the specific SK inhibitor 2-(p-hydroxyanilino)-4-(p-chlorophenyl) thi-azole, (2 mM, 72 h), led to significant inhibition (p,0.01) of DNC8 proliferation both in the absence, and presence of RA (Fig. 8B) as it did in MDA-MB-231 cells (Fig. 7). Finally, by labeling experiment with [3-3 H] D-erythro-sphingosine, we observed a significant (p,0.05) increase of intracellular S1P in DNC8 cells in response to RA (1 mM, 72 h) (Fig. 8C).
We conclude that when RA is not channeled through RARa there is no longer concerted transcriptional upregulation of nSMase-mediated CER and transcriptional downregulation of SK1 activity. In contrast, RA, through alternate, non-RAR target(s), manages to activate the SK1-S1P signaling, thus promoting cell survival and growth (Fig. 8D).

DISCUSSION
In this study we provide mechanistic evidence that RA can act as a growth inhibitor or a growth promoter according to the functional status of RARa. Moreover, we provide evidence that in the presence of a functional RARa, RA inhibits cell growth by concertedly, and inversely, regulating the synthesis of two bioactive sphingolipids, CER and S1P. In contrast, we show that in the absence of RARa, RA, in a non-RAR-mediated fashion, promotes cell growth by activating the SK1-S1P-signaling. Specifically, we found that RA, when channeled through RARa in RA-sensitive cancer cells, concertedly upregulates on one hand nSMase, thus leading to accumulation of CER, the antiproliferative and proapoptotic sphingolipid, and on the other hand downregulates SK1, pivotal for the synthesis of the oncogenic S1P, the prosurvival sphingolipid. This regulation is lost in cells (T47D) where we stably inhibited RARa function with either a RARa antagonist, or a dominant negative RARa mutant and in cells (MDA-MB-231) that lack endogenous RARa function.
Lack of RA-RARa-mediated control of the sphingolipid rheostat explains why cells survive in the presence of RA. However, we found that cells not only survive, but actually grow more in the presence of RA. Thus, RA exerts a distinct effect that is non-RAR-mediated because lack of RARa determines the downregulation/silencing of the other two RARs, RARb and RARc (data not shown). The growth-promoting action of RA and its dietary precursors has puzzled investigators for many years. Beta-carotene was shown to increase, rather than reduce, the incidence of lung cancer [19,22] and head and neck cancer [36]. Both retinol and RA were shown to promote tumor growth in transgenic models of both breast and colon cancer [18,21,35]. Here we show that chronic treatment with RA stimulates the growth of cells with an impaired RARa function not only in vitro but also in vivo. Thus, RA-induced expansion of cells may represent a discrete step of the tumor progression process once cells have lost RAR function. In the absence of functional RARs, RA apparently activates, through non-RAR targets, one or more pro-proliferative mechanisms. We provide evidence that one of these mechanisms is the SK1-S1P signaling. We do not know yet through which alternate, non-RAR target RA accomplishes to activate the SK1- S1P signaling. A potential non-RAR candidate target is protein kinase alpha (PKCa). PKCa can physically bind RA [37], can activate the SK signaling [38], and promote cell growth and tumor progression [39]. Another non-RAR candidate target is PPARb/ d, an orphan nuclear receptor that binds with high affinity RA, recently implicated in RA-induced survival [35]. PKCa and PPARb/d are both expressed in our cell model with functionally disrupted RARa (data not shown).
Our study indicates that drugs such as fenretinide that can increase CER through pathways different from the RA-RARaregulated nSMase pathway, or SK inhibitors can overcome the biological sequelae associated with the loss of RARa function and counteract RA-induced growth by targeting the sphingolipid rheostat. The identification of both non-RAR targets and mechanisms implicated in RA-mediated prosurvival/proliferation effects might bring us a step closer to the solution of the RA-paradox.

Cell cultures and biological assays
The T47D breast cancer cell line (ATCC, Manassas, VA) was cultured in DMEM medium (Invitrogen, Carlsbad, CA) plus 5% charcoal-dextran-stripped fetal bovine serum (Hyclone, Logan, UT). The T47D-derived clones, DNC8 and LXC5, carrying either a retroviral vector containing the human dominant-negative RARa403 mutant [40], or the cognate empty vector were developed as previously described [13]. The T47D-derived clone ER-C4 was developed by isolating and expanding single colonies that grew after treatment with RA 1 mM in combination with the RARa antagonist ER50891 as previously described [13]. Treatment with all-trans retinoic acid (RA) (Sigma, St Louis, MI), N-(4hydroxyphenyl) retinamide (4-HPR) (Sigma), the RARa antagonists ER50891 (provided by Dr. Kouichi Kikuchi, Discovery Research Laboratories, Ibaraki, Japan) and RO415253 (provided by Dr. Salvatore Toma, Genoa, Italy), the sphingosine kinase inhibitor, 2-(p-hydroxyanilino)-4-(p-chlorophenyl) thiazole (Calbiochem, San Diego, CA), and the neutral sphingomyelinase inhibitor (GW4869) (Sigma) are described in detail in the Results. Cell proliferation was evaluated by either the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay [41] or the  Live/Dead Cell Viability assay (Invitrogen). For the colony formation assay, exponentially growing cells were seeded at 5610 2 cells/well in 6-well plates in triplicate and allowed to attach to the substrate. Cells were treated with or without RA 1 mM for 24 hours, and then the medium was replaced with drugfree medium and the cells grown for 14-21 days. Colonies were fixed with methanol and stained with Giemsa (Sigma). The total area of the colonies was assessed by using Image J (NIH).

Animal studies and tumor analysis
Female athymic NCr-nu/nu mice (6-8 weeks old) were bought from NCI-Frederick Animal Production Program (NCI, Frederick, MD). All mice were kept in a temperature-controlled room on Figure 8. SK1-S1P signaling: a candidate growth promoting mechanism of non-RAR-mediated RA action. A) Transient exogenous expression of SK1 in DNC8 cells leads to upregulation of cyclin D1 transcription relative to cells expressing the cognate empty vector (left). Conversely, transient exogenous expression of a dominant negative SK mutant in DNC8 cells negatively affects RA-induced cyclin D1 transcription (right). B) RA-induced DNC8 proliferation is significantly decreased by treatment with a SK inhibitor. C) RA upregulates the S1P level (spots in the upper insert) in DNC8 cells. D) Scheme showing that RA action mediated through RARa results in upregulation of nSMase-generated CER sythesis, concomitant with downregulation of SK1 transcription/activity. These concerted antiproliferative metabolic changes concur to inhibit cell proliferation. Consequent to an impaired RA-RARa signaling, these concerted antiproliferative metabolic changes do not occur, thus cells survive in the presence of RA. Moreover, RA, through alternate, non-RAR (genomic or non-genomic) target(s), activates pro-survival signaling pathways, including the SK signaling pathway, thus leading to the expansion of the RA-resistant cell pool. doi:10.1371/journal.pone.0000836.g008 a 12/12-h light/dark schedule, with food and water ad libitum. Mice were estrogenized by intramuscular injection of Depoestradiol (Florida Infusion Co, Palm Harbor, FL) at 1.5 mg/kg body weight. Two days after, mice were subcutaneously inoculated in the flank region (bilaterally) with either 5610 6 DNC8 cells (16 mice) or 5610 6 LXC5 cells (16 mice) in 0.2 ml of a mixture of serum-free DMEM (Invitrogen) and Matrigel (BD Biosciences, Bedford, MA) (1:1, by vol). Mice inoculated with either LXC5 or DNC8 cells were randomly divided into two groups of 8 mice each. When mice developed palpable tumors (approximate tumor size of 20 mm 3 ), they were treated with either the vehicle, dimethylsufoxide (DMSO), or RA (2.5 mg/kg body weight) by intraperitoneal (i.p.) injection five times a week, up to six weeks. Tumors size was measured with a digital caliper twice a week, and tumor volumes were calculated as described [43]. Mice were monitored and weighed weekly. At the end of the sixth week, mice were euthanized. Data were analyzed by one-way ANOVA, followed by multiple comparison tests (STATISTICA program, Tulsa, OK, USA). All statistical tests were two-sided. The level of significance was set at p,0.05.
Three tumors (right side) randomly selected from each group of mice were removed and cut in half. One half was snap-frozen in liquid nitrogen and used for evaluating cyclin D1 transcription by quantitative real time RT-PCR. The other half was fixed in 10% neutral-buffered formalin and used for immunohistochemical analysis of the Ki67, a parameter of cell proliferation. Fixed tissues embedded in paraffin were sectioned. 5 mm-sections were reacted with either a rabbit anti-Ki67 human antibody (Dako, Carpinteria, CA) or horse serum as a control, followed by a biotinylated horse anti-rabbit antibody (BioGenex, San Ramon, CA), and visualized by using streptavidin horseradish peroxidase/ diaminobenzidine. Sections were counterstained with hematoxylin and mounted. Ki67-positive cells were quantified as described [44]. The differences between two-pair samples were analyzed by the two-tailed Student's t test.

Serine palmitoyltransferase (SPT) activity assay
The activity of SPT was determined essentially as described [50]. Briefly, 1610 7 cells were lysed by three freeze-thawing cycles in 300 ml of a lysis buffer containing 25 mM Hepes (pH 7.4), 5 mM EGTA, 50 mM NaF, 3 ml of a protease inhibitor cocktail (Sigma). The cell lysate was centrifuged at 10006g for 15 min. The supernatant was collected and protein content determined by Comassie Plus assay (Pierce Biotechnology, Inc.). The protein concentration was adjusted to 5 mg/ml with lysis buffer. 200 mg proteins were added to 160 ml solution containing 100 mM Hepes (pH 8.3), 2.5 mM EDTA, pH 7.4, 50 mM pyridoxalphosphate, 5 mM dithiothreitol, 1 mM L-serine (200 ml final volume). After a pre-incubation at 37uC for 5 min, 1 mCi L-[ 3 H(G)]serine (26.0 Ci/ mmol) (Perkin Elmer) and 20 ml 2 mM palmitoyl CoA were added. Incubation was allowed to proceed for 20 min at 37uC and stopped by adding 1.5 ml of chloroform/methanol (1/2, by vol), 25 mg Derythro-sphingosine (Avanti Polar), 1.5 ml chloroform and 2 ml 0.5 N NH 4 OH. Phases were separated by centrifugation at 20006g for 5 min. The lower organic phase was washed twice with 2 ml distilled water. Aliquots were dried and counted by liquid scintillation.

Statistical analysis
Unless specifically stated, data represent the mean of three independent experiments6standard deviation (SD). The significance of differences between groups was obtained by the Student's t-test. In the Figures one asterisk correspond to p,0.05, two asterisks to p,0.01 and three asterisks to p,0.001.