¹Baylor Breast Center, Baylor Medical College, One Baylor Plaza, MS600, Houston, Texas, TX 77030, USA

²Bristol-Myers-Squibb, Princeton, New Jersey, USA

Correspondence to: P Brown, Baylor Breast Center, Baylor Medical College, One Baylor Plaza, MS600, Houston, Texas, TX 77030, USA. E-mail: pbrown@breastcenter.tmc.edu

^aCurrent address: Department of Surgery, University of Texas Health Science Center at San Antonio, San Antonio, Texas, TX 78284, USA

We have previously shown that a retinoic acid receptor (RAR) antagonist BMS453, which does not activate RAR-dependent gene transcription in breast cells, inhibits normal breast cell growth. In this study we have investigated the mechanisms by which this retinoid receptor antagonist inhibits cell growth. Both all trans retinoic acid (atRA) and BMS453 inhibited the proliferation of normal breast cell growth without significantly inducing apoptosis. Both retinoids caused a G1 block in the cell cycle with an increase in the proportion of cells in G0/G1 and a decrease in the proportion of cells in S phase. We then investigated the effects of the retinoids on molecules that regulate the G1 to S transition. These studies demonstrated that both atRA and BMS453 induce Rb hypophosphorylation and decrease CDK2 kinase activity. We then studied the effect of the retinoids on the expression of CDK inhibitors. atRA and BMS453 increased total p21 protein levels and CDK2-bound p21 protein, but did not change CDK4-bound p21. These results suggest that atRA and BMS453 increase p21, decrease CDK2 kinase activity, which in turn leads to hypophosphorylation of Rb and G1 arrest. Because transforming growth factor beta (TGF) has been proposed as a mediator of retinoid-induced growth inhibition, we next investigated whether TGF mediates the anti-proliferative effect of atRA and BMS453 in normal breast cells. These studies showed that atRA and BMS453 increased total TGF activity by 3-5-fold. However, BMS453 increased active TGF activity by 33-fold while atRA increased active TGF activity by only threefold. These results suggest that BMS453 treatment induces conversion of latent TGF to active TGF. To investigate whether this increase in active TGF mediates the anti-proliferative effects of these retinoids, a TGF-blocking antibody was used in an attempt to prevent retinoid-induced growth inhibition. Results from these experiments showed that the anti-TGF antibody prevented the inhibition of cell proliferation induced by BMS453, but did not prevent the inhibition of cell proliferation induced by atRA. These results demonstrate that BMS453 inhibits breast cell growth predominantly through the induction of active TGF, while atRA inhibits growth through other mechanisms. These results suggest that retinoid analogs that increase active TGF may be promising agents for the prevention of breast cancer. Oncogene (2001) 20, 8025-8035.

retinoic acid; cell cycle; cyclin-dependent kinase (CDK); CDK inhibitors and normal breast cells; TGF; proliferation

atRA, all trans retinoic acid; 9cRA, 9-cis-retinoic acid; 13cRA, 13 cis retinoic acid; RAR, retinoic acid receptor; RXR, retinoic acid X receptor; ER, estrogen receptor, CDK, cyclin dependent kinase; TGF, transforming growth factor beta; HMEC, human mammary epithelial cells.

Introduction

Retinoids have been shown to prevent tumorigenesis in animals and to inhibit cancer cell growth in vivo (Roberts and Sporn, 1984). Such results led to the testing of all trans retinoic acid (atRA), 13 cis retinoic acid (13cRA), 9 cis retinoic acid (9cRA) in clinical trials for cancer treatment and prevention (Cassidy et al., 1982; Lee et al., 1993; Miller et al., 1996; Sutton et al., 1997). These trials showed that retinoids are effective for prevention and treatment of head and neck cancer and skin cancer. Thus, retinoids are most promising agents for cancer prevention. However, the toxicity of retinoids has limited their general use for cancer prevention. Therefore, current efforts are aimed at developing synthetic retinoids which are less toxic and more effective as chemoprevention agents.

Retinoids have been shown to inhibit breast cancer cell growth (Roberts and Sporn, 1984). However, the growth inhibition mechanism is not fully understood. Previous studies have shown that retinoids can inhibit breast cell proliferation (Seewaldt et al., 1997; Toma et al., 1997; Zhou et al., 1997), induce apoptosis (Fanjul et al., 1998; Toma et al., 1997, 1998a,b), or cause differentiation (Jing et al., 1996; Lee et al., 1995). Retinoids function through their receptors, the retinoic acid receptors (RAR, RAR, RAR) and the retinoid X receptors (RXR, RXR, RXR), which belong to steroid nuclear receptor superfamily (Roberts and Sporn, 1984). The expressions of retinoid receptor RAR and RAR have been shown to correlate with the antiproliferative effects of retinoids (Li et al., 1995; Liu et al., 1996; Seewaldt et al., 1995; Sheikh et al., 1994).

It has been proposed that retinoids inhibit cell growth by inhibiting the expression of growth stimulating factors or signaling pathways, or by inducing the expression of growth inhibitory factors such as TGF (Pfahl, 1993; Yang-Yen et al., 1991). Retinoids induce the secretion of TGF in many different cells including breast cancer cells (Glick et al., 1989; Pfahl, 1993; Yang-Yen et al., 1991). In HL-60 cells, the induction of TGF has been shown to be necessary for retinoid-induced growth inhibition of these cells (Nunes et al., 1996). TGF proteins are secreted in a latent inactive form, in which TGF is noncovalently associated with latency associated peptide (LAP). Latent TGF is converted to an active form when LAP is removed from the complex by proteolysis through multiple pathways (Gleizes et al., 1997). Active TGF functions through binding to TGF receptors, which signal through Smad signal transduction pathways (Lagna et al., 1996; Liu et al., 1998).

Retinoids have also been shown to inhibit breast cell proliferation by causing G1 arrest. Several cell cycle modulators are important to control G1 transition to S phase include Rb, cyclins D and E, cyclin-dependent kinases 2, 4, and 6, and their inhibitors, p15, p16, p21, and p27 (Hunter and Pines, 1994; Sher, 1994). Retinoids have been shown to induce Rb dephosphorylation (Seewaldt et al., 1997; Zhu et al., 1997) and reduction of cyclin D1, D3, and E protein levels in normal and cancerous breast cells and immortalized human bronchial epithelial cells (Langenfeld et al., 1996; Zhou et al., 1997). atRA has been reported to decrease CDK2 and CDK4 kinase activities in breast cancer cells (Teixeira and Pratt, 1997; Zhou et al., 1997). Retinoids have been shown to affect p21 expression in breast cancer cells. atRA has been reported to decrease p21 protein level in breast cancer cells (Zhu et al., 1997), or have no effect on p21 protein level in normal breast cells (Seewaldt et al., 1997). The synthetic retinoid, CD437 (AHPN), a RAR gamma-selective retinoid, has been shown to increase the expression of p21 protein in breast cancer cells through a mechanism independent of RAR and RXR (Li et al., 1996; Shao et al., 1995). Recent studies by Zhang et al. (2000) have also shown that CD437 increases p21 in normal mammary epithelial cells.

Naturally occurring retinoids are promising agents for the prevention of epithelial cancers, however, their toxicity has limited their general use. By elucidating the mechanism by which retinoids inhibit growth, synthetic retinoids that retain chemopreventive efficacy but that have reduced toxicity can be developed. We have previously shown that a synthetic retinoid BMS453, an RAR antagonist that binds RAR proteins, but does not activate RAR-dependent gene expression, inhibited breast cell growth (Yang et al., 1999). In these studies we have shown that BMS453 inhibits cell proliferation without inducing apoptosis in normal breast cells. We further characterized the effects of BMS453 and atRA on the cell cycle and cell cycle modulators in normal breast cells. These results demonstrate that both atRA and BMS453 caused G1 arrest. Treatment with atRA and BMS453 caused increased expression of p21 protein, decreased CDK2 kinase activity, and Rb hypophosphorylation. Since BMS453 is an RAR antagonist, and does not appear to activate RAR-dependent gene expression, the increase in p21 protein expression is likely independent of RAR-dependent transactivation. The present results also show that BMS453 is a more potent activator of TGF than atRA, and that the activation of TGF is necessary for BMS453-induced growth inhibition of normal breast cells. These studies demonstrate that it is possible to identify retinoid analogs that inhibit cell growth without activating typical RAR-dependent pathways. Such agents will likely have less toxicity than naturally occurring retinoids, and may be useful agents for breast cancer prevention.

Results

atRA and BMS453 inhibit ³H-thymidine incorporation in normal breast cells

We have previously shown that atRA and BMS453 inhibit normal and malignant breast cells, and that this inhibition does not necessarily depend on RAR-induced transactivation (Yang et al., 1999). In the present study we investigated the mechanism by which atRA and BMS453 inhibit normal breast cell growth. In order to investigate whether retinoids inhibit normal breast cell proliferation, we used a ³H-thymidine incorporation assay to study the effects of retinoids on DNA synthesis. As seen in Figure 1, both atRA (1 M) and the synthetic retinoid (BMS453) (1 M) inhibited ³H-thymidine uptake in normal breast cells (184 and HMEC) by 40-60%. Therefore, both atRA and BMS453 inhibited cell proliferation in normal breast cells.

Retinoids do not induce apoptosis in normal breast cells

Next, we studied whether retinoid-induced growth inhibition was due to apoptosis in normal breast cells. Using a TUNEL assay, 1 M atRA induced a slight increase in apoptosis in HMEC, which was not statistically significant, while 1 M BMS453 did not induce apoptosis (Figure 2a,b). Taxol, an apoptosis-inducing agent (Gangemi et al., 1996), did cause a low level of apoptosis in normal breast cells (Figure 2a,b), although these normal cells are less resistive to Taxol than are cancer cells. Using this assay, breast cancer cells show >50% apoptosis when treated with Taxol (data not shown). A second measure of apoptosis, assessing cytoplasmic fragmentation of DNA, was performed to further monitor sub-microscopic apoptotic events. As shown in Figure 2c, 1 M atRA and 1 M BMS453 did not significantly induce apoptosis as measured by this assay.

atRA and BMS453 cause G1 arrest in normal breast cells

To investigate the mechanism by which BMS453 inhibits proliferation, we studied the effect of retinoids on cell cycle. Normal breast cells (184 cells) were treated with vehicle (DMSO), 1 M atRA and 1 M BMS453 for 5 days and flow cytometry assay was performed. The results of a representative experiment from at least three independent experiments are shown in Figure 3. One M atRA and 1 M BMS453 increased the proportion of cells in G0/G1 phase (by 11-37%) and decreased the proportion of cells in S phase (by 18-63%). This indicates that both atRA and BMS453 inhibit normal breast cell proliferation by causing G1 arrest.

Effects of atRA and BMS453 on Rb phosphorylation, cyclins and CDK expression

Molecules that regulate G1 to S transition include Rb, cyclins D and E, CDK2, CDK4, and the CDK inhibitors. We first investigated the effects of retinoids on the phosphorylation of Rb. As seen in Figure 4 (top panel), Rb is highly phosphorylated in exponentially growing cells (lane 2) and vehicle (DMSO)-treated cells (lanes 4, 7 and 10). Rb is hypophosphorylated in cells treated with TGF or by starving of growth factors (lanes 1 and 3). Both atRA and BMS453 also induced Rb hypophosphorylation after 48 and 73 h (lanes 7-9 and lanes 10-12).

Since Rb is phosphorylated by cyclin E : CDK2 or cyclin D : CDK4/6 complexes, we next determined the effects of retinoids on protein levels of cyclin D and cyclin E. As shown in Figure 4, cyclin D1 and cyclin E protein levels were dramatically decreased in starved cells (lane 3). In cells treated with atRA or BMS453, cyclin D1 protein expression was unchanged. Cyclin E protein expression was also unchanged in cells treated with atRA or BMS453 for 24 to 48 h (lanes 4-6 and lanes 7-9), but was reduced in cells treated with these retinoids after 72 h (lanes 10-12).

We next determined the effects of retinoids on the expression of CDK2 and CDK4 proteins. As seen in Figure 4, CDK2 and CDK4 expression decreased in starving cells (lane 3). The expression of CDK2 was also reduced by atRA after 48 and 72 h (Figure 4, lanes 7-9 and lanes 10-12). CDK2 showed the greatest decrease after 72 h of treatment with atRA.

atRA and BMS453 decrease CDK2 kinase activity

Since both cyclin D : CDK4/6 and cyclin E : CDK2 complexes phosphorylate Rb, we next investigated the effects of retinoids on cyclin-associated kinase activities. Histone H1 was used as a substrate for detecting CDK2 kinase activity while GST-Rb was used as a substrate for detecting CDK4/6 kinase activity (cyclin D-associated kinase activity). As seen in Figure 5, proliferating cells had much higher CDK2 kinase activity (lane 1) as compared with starved cells or TGF-treated cells (lanes 2 and 9). One M atRA and 1 M BMS453 decreased CDK2 kinase activity by 40-50% as shown by decreased phosphorylation of histone H1 after treatment for 24 and 72 h (lanes 3-5 and lanes 6-8). However, CDK4/6 kinase (cyclin D-associated kinase) activity was not affected by retinoids. Since cyclin D-associated kinase activity contains both CDK4 and CDK6 kinase activities, we also determined the effect of retinoid treatment on CDK4 kinase activity (using a CDK4 antibody to immunoprecipitate the CDK4 complex, and then using GST-Rb as the substrate for CDK4 kinase). Neither atRA nor BMS453 decreased CDK4 kinase activity after treatment for 24 and 72 h (data not shown). Therefore, the hypophosphorylation of Rb induced by these retinoids is due to the decreased CDK2 kinase activity in retinoid-treated breast cells.

Retinoids increase p21, p15, CDK2-associated p21 and CDK4-associated p15 proteins

We next determined the effects of retinoids on protein expression of CDK inhibitors such as p27, p21, p16 and p15. p27 protein was not detectable in normal breast cells (data not shown). As shown in Figure 6, p21 expression was increased by TGF treatment (lane 1). P21 was also increased by both atRA and BMS453 after 24 h (lanes 4-6). P21 expression returned to baseline after 48 h (lanes 7-9), and then decreased after 72 h (lanes 10-12) treatment with the retinoids. Neither retinoid significantly affected the p16 protein level (lanes 4-6 and lanes 7-9), except after treatment for 72 h, when p16 expression was decreased by atRA but not by BMS453 (lanes 10-12). p15 was increased by TGF treatment and starving (lanes 1-3), and by atRA and BMS453 treatment after 24 and 48 h (lanes 4-6 and lanes 7-9), but was not significantly changed after 72 h (lanes 10-12).

We next determined the levels of p21 and p15 proteins present in the CDK2 and CDK4 complexes by first immunoprecipitating the CDK2 or CDK4 complexes with specific antibodies, and then performing by Western blot analysis for the p21 and p15 proteins. As shown in Figure 7, after treatment for 24 h with the retinoids, the level of CDK2-associated p21 protein level was increased (lanes 1-3). Relative levels of CDK2-associated p21 were determined by densitometric quantitation followed by normalization to total CDK2 levels. These results show that atRA and BMS453 increase CDK2-associated p21 by 3.1 and 2.3-fold, respectively after 24 h treatment. Neither retinoid increased CDK4-associated p21 (lanes 7-12). CDK4-associated p15 protein levels were increased by atRA and BMS453 after 72 h (by 2.0-fold by atRA and 2.2-fold by BMS453, lanes 10-12). These results are consistent with the Western blot analysis shown in Figure 6 demonstrating that the retinoids increased total p21 and p15.

atRA and BMS453 both induce TGF activity in normal breast cells

Because some of the antiproliferative effects of retinoids have been linked to the production of TGF (Glick et al., 1989; Nunes et al., 1996), we next investigated the effect of these retinoids on TGF expression and activity. Normal human mammary epithelial cells (184) were treated with 1 M atRA and 1 M BMS453, and RNAs were prepared from these retinoid-treated cells. RNase protection assays were then performed to measure TGF 1, 2, and 3 mRNA expression. One M atRA and 1 M BMS453 has no significant effect on mRNA level of TGF1, TGF2 in normal breast cells, while TGF3 was not detected in these cells (data not shown). We also investigated the effect of the retinoids on the expression of the receptors for TGF, TGF receptors I and II. Using an RNase protection assay we found that the 184 normal breast cells express TGF receptors I and II, and that 1 M atRA or BMS453 did not affect their RNA expression levels (data not shown).

TGF activity was then measured by incubating the medium from 184 cells treated with retinoids (conditioned medium) with the mink lung epithelial cells, which have been stably transfected with PAI-1 promoter luciferase construct containing a TGF response element (Abe et al., 1994). By heating the 184-conditioned media, latent TGF can be converted to active TGF. Thus, in heated samples, TGF activity detected by this assay represents the active TGF plus latent TGF in the original conditioned media (referred to as total TGF). As shown in Figure 8, 1 M atRA increased total TGF activity by twofold, fivefold and 5.5-fold after 1, 3, and 5 days of treatment respectively (P values less than 0.05, 0.001, 0.05 respectively), while 1 M BMS453 caused a slight increase in total TGF (not statistically significant). As described in Materials and methods, by using conditioned medium that is not heated, this assay measures the active TGF present in the conditioned media, without measuring the latent TGF. Therefore, we also measured only the active TGF in the conditioned media from cells treated with atRA or BMS453. As shown in Figure 8, 1 M atRA increased active TGF activity by twofold and fivefold after 3 and 5 days of treatment (P values less than 0.05, 0.05 respectively), while 1 M BMS453 increased active TGF activity by 4- and 33-fold after 3 and 5 days of treatment (P values less than 0.05, 0.005 respectively). Thus, these results demonstrate that both atRA and BMS453 increase the amount of total TGF. However, BMS453 induces much more active TGF than does atRA. This is particularly striking given the fact that atRA induces more total TGF than BMS453. These results suggest that treatment with BMS453 leads to activation of latent TGF into active TGF.

TGF blocking antibody prevents BMS453-induced inhibition of normal breast cell growth

Since atRA and BMS453 increased TGF activity, we next determined whether TGF is required for the growth inhibition induced by these retinoids using a TGF blocking antibody. For these experiments, uptake of ³H-thymidine was measured in breast cells treated with retinoids, vehicle alone, or TGF1, in the absence or presence of this TGF-blocking antibody. As shown in Figure 9, cells treated with vehicle (DMSO) plus IgG, there was a high level of ³H-thymidine uptake, and the TGF-blocking antibody did not inhibit ³H-thymidine uptake (bars 1 and 2). However, TGF does inhibit ³H-thymidine uptake of these cells, and the TGF-blocking antibody reversed this inhibitory effect of TGF (bars 3 and 4). atRA suppressed ³H-thymidine incorporation either in the presence of IgG or the TGF-blocking antibody (bars 5 and 6). However, while 1 M BMS453 also inhibited ³H-thymidine incorporation in the presence of IgG, the TGF-blocking antibody was able to reverse this inhibitory effect of BMS453 (bars 7 and 8). Therefore, these data demonstrate that TGF mediates the anti-proliferative effect of BMS453 but not of atRA in normal human mammary epithelial cells.

Discussion

The above results demonstrate that both atRA and the synthetic retinoid BMS453 inhibit normal mammary epithelial cell growth by causing G1 arrest without significantly inducing apoptosis. The results presented here show that treatment of these cells with these retinoids increases p21 expression, which inhibits CDK2 kinase activity and causes Rb hypophosphorylation. Both of these retinoids also induce TGF expression. However, unlike atRA, BMS453 strongly induces active TGF. A TGF-blocking antibody totally reversed the growth suppression induced by BMS453, demonstrating that the anti-proliferative effect of BMS453 is mediated by TGF.

Retinoids inhibit breast cell growth by several pathways. We and others have shown that retinoids inhibit the AP-1 transcription factor through transcription factor crosstalk (Yang et al., 1999; Chen et al., 1995; Lin et al., 2000). Both atRA and BMS453 have been shown to inhibit AP-1 transcription factor (Yang et al., 1999; Chen et al., 1995). Further studies on the role of AP-1 transrepression in retinoid-induced growth suppression are ongoing in our laboratory.

Retinoids have also been shown to cause G1 cell cycle arrest (Seewaldt et al., 1997), and apoptosis of breast cancer cells (Toma et al., 1997) depending on the specific retinoid tested. The present results demonstrate that atRA and BMS453 do not significantly induce apoptosis in normal human mammary epithelial cells but instead inhibit their proliferation. Our results are consistent with the study by Seewaldt et al. (1997) that showed that atRA causes G1 arrest without inducing apoptosis in normal breast cells.

Cyclin E-CDK2 complex and cyclin D-CDK4 are necessary to govern G1 to S phase transition. Langenfeld et al. (1996) have shown that retinoids downregulated cyclin D and cyclin E protein levels in immortalized human bronchial epithelial cells. Seewaldt et al. (1997) has shown that atRA decreased cyclin D1 protein and cyclin E protein in normal breast cells. Our results showed that atRA and BMS453 did not significantly change the protein level of cyclin D1, but did reduce cyclin E expression in HMEC and 184 normal breast cells. These differences may be due to the fact that we used different isolates of human mammary epithelial cells than did Seewaldt's group (Seewaldt et al., 1997). We also found that atRA and BMS453 decreased CDK2 activity (but not cyclin D-associated (CDK4/6) or CDK4 kinase activity) in normal breast cells. Therefore, it is likely that decreased CDK2 kinase activity leads to hypophosphorylation of Rb. These results are consistent with previous studies by Teixeira and Pratt (1997), which showed that CDK2 was a target for atRA-induced G1 arrest in MCF7 cells.

Since CDK2 kinase activity is inhibited by p21, we also examined the effects of atRA and BMS453 on p21 expression. The effects of the retinoids on p21 protein have been studied by several groups. Retinoids have been reported to increase p21 protein expression in some human breast cells (Li et al., 1996; Shao et al., 1995; Zhang et al., 2000), to decrease p21 expression in MCF7 cells (Zhou et al., 1997; Zhu et al., 1997), or to not change p21 expression in normal breast cells (Seewaldt et al., 1997). Our results showed that atRA and BMS453 increased p21 protein levels in HMEC and 184 normal breast cells after 24 h. AtRA may induce p21 expression directly though activation of the retinoic acid response element (RARE) within the p21 promoter. However, BMS453 does not activate RARE-dependent gene expression, so it likely increases p21 expression through some other mechanism. Since atRA and BMS453 also increased the level of total or active TGF, this increase in TGF also likely induces p21 expression, which in turn reduces CDK2 activity, and causes Rb hypophosphorylation and a G1 cell cycle block. It should be noted that the effects of the retinoids on these cell cycle regulators are less dramatic than seen in cells treated with purified TGF. This is likely because the concentration of the added TGF is much higher than the concentration of the TGF elicited by retinoids.

Regulation of TGF gene expression by retinoids may be at the post-transcriptional level since there is no RARE identified in the promoter of TGF (Kim et al., 1989; Lafyatis et al., 1990, Noma et al., 1991). The observation that atRA and BMS453 did not increase TGF mRNA level but did increase TGF activity is consistent with the hypothesis that these retinoids increase TGF activity through a post-transcriptional mechanism. Our results suggest that treatment with BMS453 somehow activates the conversion of latent TGF into active TGF. Multiple factors such as plasmin, mannose 6-phosphate receptor/IGF-II receptor, tissue transglutaminase or cathepsin D are involved in the conversion of latent TGF into active TGF (Glick et al., 1989). Previous studies indicate that retinoids can activate the conversion of latent TGF into active TGF by increasing plasminogen activator level (Imai et al., 1997; Gleizes et al., 1997; Kojima and Rifkin, 1993). Studies to investigate how BMS453 activates latent TGF are ongoing. The mechanism by which atRA inhibits normal breast cell growth is likely to be more complex. Of note, the present results demonstrate that block of the TGF is not sufficient to reverse the anti-proliferative effect of atRA, indicating other pathways transduce some of the growth suppressive signals induced by atRA.

Based on the above results, we propose a model for the mechanism by which atRA and BMS453 cause G1 arrest (shown in Figure 10). As seen in this figure, BMS453 induces TGF activation through an unknown mechanism that is likely independent of RAR-dependent transactivation. Active TGF signals through Smads to increase the p21 protein level, which then inhibits CDK2 activity, causing Rb hypophosphorylation and G1 arrest. As we have previously shown, BMS453 also inhibits the AP-1 transcription factor through transcription factor crosstalk (Yang et al., 1999). This inhibition of AP-1 may somehow lead to activation of latent TGF, or alternatively could alter TGF signaling to increase the sensitivity of breast cells to the antiproliferative effects of TGF. In contrast, atRA likely increases p21 expression through another mechanism, such as by direct activation of the RARE within the p21 promoter, which also leads to G1 arrest.

Retinoids are promising chemopreventive agents, however naturally occurring retinoids are likely too toxic for general use. The synthetic retinoid BMD453, which does not activate RAR-dependent transcription, but which does inhibit normal breast cell growth by inducing active TGF, represents a class of synthetic retinoids that may retain their chemopreventive efficacy, yet have less toxicity. Future studies will focus on determining its effect on tumorigenesis, the toxicity and chemopreventive efficacy of this and other synthetic retinoids using breast cancer animal models. Results from these studies should lead to the development of effective and safe agents for the prevention of breast cancer.

Materials and methods

Cell lines and reagents

Normal human mammary epithelial cells were grown in mammary epithelium basal medium (MEBM) supplemented with bovine pituitary extract, human epidermal growth factor, insulin, hydrocortisone, gentamicin sulfate-100 (Clontech Laboratories, Inc., Palo Alto, CA, USA). all trans retinoic acid was obtained from Sigma. The retinoic acid antagonist BMS453 was obtained from Bristol-Mayers-Squibb. All experiments with retinoids were performed in reduced light. Antibodies for cyclin E, CDK2, CDK4, p21, p16 and p15 were purchased from Santa Cruz Biotechnology Inc. (Santa Cruz, CA, USA). Antibody for cyclin D1 was purchased from PharMingen (San Diego, CA, USA). Rb antibody is kindly provided by Dr Wen-Hwa Lee (Institute of Biotechnology, University of Texas Health Science Center at San Antonio, TX, USA). Anti-EGF receptor antibody was purchased from UBI (Lake Placid, NY, USA), and neutralizing TGF antibody was purchased from RD (AB-100NA, Minneapolis, MN, USA).

³H-Thymidine incorporation assay

Fifty thousand cells were seeded in a 24-well plate and then were starved in medium without any growth factors to synchronize the cells. After 24 h of growth factor deprivation, the cells were treated with retinoids for 11 h in medium containing pituitary extract and other growth factors. The cells were labeled with ³H-thymidine (2 Ci/ml) for 3 h and then were incubated with 5% TCA at 4°C for 30 min. The cells were then lysed by addition of 0.1 N NaOH. The level of protein in the lysates was determined using a BCA assay (Pierce, Rockford, IL, USA). ³H-Thymidine uptake was measured by mixing the cell lysates with scintillation fluid and counting the ³H c.p.m. in a scintillation counter. Each data point was performed in triplet, and the results were reported as c.p.m.±standard error. All results were normalized to protein content.

Apoptosis assay

For TUNEL assay, eight thousand cells were seeded in 8-chamber slides and then were treated with retinoids (1 M) for 5 days. The retinoids were added to the cells every other day. TUNEL assays were performed according to TACS^TM in situ Apoptosis Detection Kit (Trevigen, Gaithersburg, MD, USA). The number of apoptotic cells (dark brown cells) and total number of cells were counted, and the results were expressed as number of apoptotic cells per 1000 cells. For cell death ELISA assay, 50 000 cells were seeded in a 6-well plate and treated with retinoids (1 M) for 6 days. The apoptosis assay was performed according to Cell Death Detection ELISA kit (Boehringer Mannheim). Each sample was performed in triplet and the results expressed as mean±standard error from at least two independent experiments.

RNA extraction and RNase protection assay

RNA from breast cells was prepared by lysis of cells in GIT buffer as previously described (Chirgwin et al., 1979). RNase protection assays were performed as described by Rocha et al. (1996). In brief, ³²P-UTP labeled antisense cRNA was transcribed from human TGF1, 2, 3, and 36B4 cDNAs by RNA polymerase as described (Rocha et al., 1996; Sun et al., 1994). Digestion of the cRNA : mRNA hybrids produced 243, 337 and 181 bp protected bands for TGF1, 2 and 3 respectively. The 36B4 riboplasmid was used to normalize for differences in loading (Rocha et al., 1996). Thirty g RNAs were hybridized with ³²P-UTP labeled antisense RNAs followed by RNase A and RNase T treatment. The hybrids were loaded on 8 M urea 6% SDS-PAGE. The gel was dried and exposed to X-ray film. The intensity of signals was quantitated by PhosphorImager (Molecular Dynamics, Sunnyvale, CA, USA).

Flow cytometry assay

Normal breast cells were treated with vehicle (DMSO) or retinoids for 5 days and were stained with propidium iodide solution containing 50 mM EDTA, propidium iodide (50 g/ml) and 1.0 mg/ml RNase A at 1.0´10⁶ cells/ml for 45 min at room temperature in the dark. Prior to flow cytometric measurements, samples were filtered through a 37 mm nylon mesh into 12´75 mm tubes and stored at 4°C until analysis was performed within 24 h. Stained cells were analysed with an EPICS ELITE flow cytometer (Beckman-Coulter, Miami, FL, USA). Histograms were analysed for cell cycle compartments using MultiCycle-PLUS Version 3.0 (Phoenix Flow Systems, San Antonio, CA, USA).

Western blot analysis

Whole cell extracts were prepared by suspending the cells in lysis buffer (50 mM Tris pH 6.8, 2% SDS) and the protein concentration was determined by BCA assay (Pierce, Rockford, IL, USA). Thirty g to 100 g of protein was fractionated on a 10-12% acrylamide denaturing gel and transferred onto a nitrocellulose membrane (Life Science, Amersham) by electroblotting. The membrane was blocked with 5% non-fat dry milk in TBST (50 mM Tris pH 7.5, 150 mM NaC1, 0.05% Tween 20) for 1 h at room temperature or overnight at 4°C and washed in TBST. The membrane was then incubated with primary antibody at a 1 : 100 to 1 : 200 dilution for 1 h at room temperature. After washing with TBST, the membrane was incubated with horseradish peroxidase-conjugated secondary antibody (Life Science, Amersham) at 1 : 4000 dilution for 1 h at room temperature. The membrane was washed in TBST and then the color was developed using the enhanced chemiluminescence (ECL) procedure (Life Science, Amersham).

Immunoprecipitation and kinase assays

Cells were starved with mammary epithelium basal medium (MEBM) (Clontech Laboratories, Inc., Palo Alto, CA, USA) with anti-human epidermal growth factor receptor antibody to effectively block autocrine TGF signals (Upstate Biotechnology, Lake Placid, NY, USA) for 2 days and then treated with retinoids in MEBM with full supplements as described above. For the CDK2 kinase assay, cells were lysed in lysis buffer containing 50 mM HEPES pH 7.5, 150 mM NaC1, 1 mM EGTA, 1 mM DTT, 1% Triton X-100, 10% glycerol, 10 mM -glycerophosphate, 100 mM NaF, 0.2 mM NaVO₃, 1.5 mM MgC1₂, 10 g/ml aprotinin, 10 g/ml leupeptin and 1 mM PMSF. For CDK4 kinase assay, cells were lysed in the lysis buffer containing 50 mM HEPES pH 7.5, 150 mM NaC1, 1 mM EDTA pH 8, 2.5 mM EGTA, 1 mM DTT, 0.1% Tween 20, 10% glycerol, 10 mM -glycerophosphate, 1 mM NaF, 0.1 mM NaVO₃, 1.5 mM MgC1₂, 2 g/ml aprotinin, 10 g/ml leupeptin and 0.1 mM PMSF. Protein concentration was determined by BCA assay (Pierce, Rockford, IL, USA). Protein G agarose (Life Technologies, Gaithersburg, MD, USA) was incubated with CDK2 or cyclin D1 antibody at 4°C for 1 h followed by incubating with 500-800 g of protein extracts at 4°C overnight. The agarose mixture was pelleted and washed in lysis buffer four times. For immunoprecipitation-Western blotting, the agarose was resuspended in 40 l 1´ sample buffer (125 mM Tris pH 6.8, 4% SDS, 0.005% bromophenol blue, 20% glycerol, 0.7 M -mercaptoethenol) and 20 l was loaded on 12% SDS-PAGE. Western blotting was performed as described above. For the CDK2 kinase assay, the agarose mixture was washed in 1´ cold kinase buffer (20 mM Tris pH 7.5, 5 mM MgC1₂, 2.5 mM MnC1₂ and 1 mM DTT) and resuspended in final volume of 25 l containing 5 l 5´ kinase buffer (100 mM Tris pH 7.5, 25 mM MgC1₂, 12.5 mM MnC1₂, 5 mM DTT), 1 l [-³²P] ATP and 20 g of histone H1 (Boehringer Mannheim, Indianapolis, IN, USA). For cyclin D-associated (CDK4/6) kinase assay, the agarose mixture was washed in 1´ cold kinase buffer (50 mM HEPES pH 7.5, 10 mM MgC1₂, 1 mM DTT, 10 mM -glycerophosphate, 1 mM NaF and 0.1 mM NaVO₃) and resuspended in final volume of 25 l containing 5 l 5´ kinase buffer (250 mM HEPES pH 7.5, 50 mM MgC1₂, 5 mM DTT, 50 mM -glycerophosphate, 5 mM NaF, 0.1 mM NaVO₃, 0.1 mM ATP), 1 l [-³²P] ATP and 1 g GST-Rb (Santa Cruz Biotechnology Inc., Santa Cruz. CA, USA). The kinase reaction was performed at 30°C for 30 min and was stopped by adding 25 l of 2´ sample buffer. The samples were heated at 90°C for 5 min and 25 l of reaction mixture was loaded on 10% SDS-PAGE. The gel was dried and exposed to X-ray film. The intensity of the bands was quantitated using a PhosphorImager (Molecular Dynamics, Sunnyvale, CA, USA).

TGF reporter assays

To measure TGF activity, a TGF-sensitive bioassay was used as previously developed by Abe et al. (1994). In this assay, active TGF is detected by the induction of luciferase activity in mink lung epithelial cells which have been stably transfected with a TGF-responsive (from PAI-1 promoter) luciferase reporter gene. Using this assay, active TGF can be detected in the conditioned media from cells grown in culture. In addition, total TGF (active plus latent TGF) secreted into the medium can also be measured after converting latent TGF into active TGF by heating the conditioned media (80°C for 10 min). 1´10⁵ 184 normal breast cells were seeded on 6-well plate and treated with atRA or BMS453 for 1, 3, or 5 days. Pituitary extract was excluded from the media for 24 h before collecting medium from the 184 cells. These conditioned media were then added to the mink lung epithelial cells (1.6´10⁵ cells per 96-well). After culturing for 16 h, the mink lung epithelial cells were then lysed in lysis buffer (1 mM DTT, 100 mM potassium phosphate pH 7.8 and 1% Triton X-100) and luciferase activity was measured (Abe et al., 1994). The data were plotted as mean±standard error from at least two independent experiments.

Acknowledgements

We would like to thank Dr Michael Brattain for providing us with the TGF1, 2, 3, TGFI and II receptor riboplasmids, Dr Doug Yee for providing 36B4 riboplasmid, Dr Rafael Herrera for technical assistance in performing CDK kinase assays, and Dr Marco Gottardis for helpful review of the manuscript. This work was supported by Cancer Research Foundation of America (LM Yang), NIH Grant R01CA78480 (P Brown), and V Foundation for Cancer Research (P Brown).

Abe M, Harpel JG, Metz CN, Nunes I, Loskutoff DJ, Rifkin DB. (1994). Analytical Biochem. 216, 276-284.

Cassidy J, Liffman M, Lacroix A, Peck G. (1982). Eur. J. Cancer Clin. Oncol. 18, 925-928.

Chen JY, Penco S, Ostrowski J, Balague P, Pon M, Starrett J, Reczek P, Chambon P, Gronemeyer H. (1995). EMBO J. 14, 1187-1197. MEDLINE

Chirgwin JM, Przybyla AE, MacDonald RJ, Rutter WJ. (1979). Biochem. 18, 5294-5299.

Fanjul AN, Piedrafita EJ, Al-Shamma H, Pfahl M. (1998). Cancer Research 58, 4607-4610.

Gangemi RM, Costa G, Fulco RA, D'aquino S, Aronadio O. (1996). Ann. NY Acad. Sci. 784, 550-554.

Gleizes P-E, Munger JS, Nunes I, Harpel JG, Mazzieri R, Noguera I, Rifkin D. (1997). Stem Cells 15, 190-197.

Glick AB, Flanders KC, Danielpour D, Yuspa SH, Sporn MB. (1989). Cell Regul. 1, 87-97.

Hunter T, Pines J. (1994). Cell 79, 573-582. MEDLINE

Imai S, Okuno M, Moriwaki H, Muto Y, Murakami K, Shudo K, Suzuki Y, Kojima S. (1997). FEBS Lett. 411, 102-106. MEDLINE

Jing Y, Zhang J, Waxman S, Mira-y-Lopez R. (1996). Differentiation 60, 109-117.

Kim SJ, Glick A, Sporn MB, Roberts AB. (1989). J. Biol. Chem. 264, 402-408.

Kojima S, Rifkin DB. (1993). J. Cell. Physiol. 155, 323-332.

Lafyatis R, Lechleider R, Kim SJ, Jakowlew S, Roberts AB, Sporn MB. (1990). J. Biol. Chem. 265, 19128-19136. MEDLINE

Lagna G, Hata A, Hemmati-Brvanlou, Massague J. (1996). Nature 383, 832-836. MEDLINE

Langenfeld J, Lonardo F, Kiyokawa H, Passalaris T, Ahn M, Rusch V, Dmitrovsky E. (1996). Oncogene 13, 1983-1990. MEDLINE

Lee JS, Newman RA, Lippman SM, Huber MH, Minor T, Raber MN, Krakoff IH, Hong WK. (1993). Journal of Clinical Oncology 11, 959-966.

Lee PT, Lee MT, Darcy KM, Shudo K, Ip MM. (1995). Endocrinology 136, 1707-1717.

Li X, Rishi A, Shao Z, Dawson M, Jong L, Shroot B, Reichert U, Ordonez J, Fontana J. (1996). Cancer Research 56, 5055-5062.

Li X-S, Shao Z-M, Sheikh MS, Eiseman JL, Sentz D, Jetten AM, Chen J-C, Dawson MI, Aisner S, Rishi AK, Gutierrez P, Schnapper L, Fontana JA. (1995). Journal of Cellular Physiology 165, 449-458.

Lin F, Xiao D, Kolluri S, Zhang X. (2000). Cancer Research 60, 3271-3280.

Liu R, Lee M-O, Wang H-G, Li Y, Hashimoto Y, Klaus M, Reed JC, Zhang X-K. (1996). Molecular and Cellular Biology 16, 1138-1149.

Liu X, Yue J, Frey RS, Zhu Q, Mulder KM. (1998). Cancer Res. 58, 4752-4757.

Miller VA, Rigas JR, Benedetti FM, Verret AL, Tong WP, Kris MG, Gill GM, Loewen GR, Truglia JA, Ulm EH, Warrell Jr RP. (1996). Clinical Cancer Research 2, 471-475.

Noma T, Glick AB, Geiser AG, O'Reilly MA, Miller J, Roberts AB, Sporn MB. (1991). Growth Factors 4, 247-255.

Nunes J, Kojima S, Rifkin DB. (1996). Cancer Res. 56, 495-499.

Pfahl M. (1993). Endocrine Rev. 14, 651-658.

Roberts AB, Sporn MB. (1984). The Retinoids. Sporn MB, Roberts AB and Goodman DS (eds). Academic Press: Orlando, FL, pp 209-286.443-520.

Rocha RL, Hilsenbeck SG, Jackson JG, Lee AV, Figueroa JA, Yee D. (1996). J. Natl. Cancer Inst. 88, 601-606.

Seewaldt VL, Johnson BS, Parker MB, Collins SJ, Swisshelm K. (1995). Cell Growth & Differ. 6, 1077-1088.

Seewaldt VL, Kim J-H, Caldwell LE, Johnson BS, Swisshelm K, Collins SJ. (1997). Cell Growth & Differ. 8, 631-641.

Shao Z-M, Dawson MI, Li XS, Rishi AK, Sheikh MS, Han Q-X, Ordonez JV, Shroot B, Fontana JA. (1995). Oncogene 11, 493-504. MEDLINE

Sheikh MS, Shao Z-M, Li X-S, Dawson M, Jetten AM, Wu S, Conley BA, Garcia M, Rochefor H, Fontana JA. (1994). The Journal of Biological Chemistry 269, 21440-21447.

Sher C. (1994). Cell 79, 551-555. MEDLINE

Sun L, Wu G, Willson JKV, Zborowska E, Yang J, Rajkarunanayake I, Wang J, Gentry LE, Wang X-F, Brattain MG. (1994). J. Biol. Chem. 269, 26449-26455. MEDLINE

Sutton LM, Warmith MA, Petros WP, Winter EP. (1997). Cancer Chemother. Pharmacol. 40, 335-341.

Teixeira C, Pratt MA. (1997). Mol. Endocrinol. 11, 1191-1202. MEDLINE

Toma S, Isnardi L, Riccardi L, Bollag W. (1998a). Anticancer Research 18, 935-942.

Toma S, Isnardi L, Raffo P, Dastoli G, DeFrancisci E, Riccardi L, Palumbo R, Bollag W. (1997). Int. J. Cancer. 70, 619-627. Article MEDLINE

Toma S, Isnardi L, Raffo P, Riccardi L, Dastoli G, Apfel C, LeMotte P, Bollag W. (1998b). Int. J. Cancer 78, 86-94. Article MEDLINE

Yang L, Munoz-Medellin D, Kim H-T, Ostrowski J, Reczek P, Brown PH. (1999). Breast Cancer Research and Treatment 56, 277-291.

Yang-Yen HF, Zhang X-K, Graupner G, Tzukerman M, Sakamoto B, Karin M, Pfahl M. (1991). The New Biologist 3, 1206-1219. MEDLINE

Zhang Y, Rishi A, Dawson M, Tschang R, Farhana L, Boyanapalli M, Reichert U, Shroot B, Buren EV, Fontana J. (2000). Cancer Research 60, 2025-2032.

Zhou Q, Stetler-Stevenson M, Steeg PS. (1997). Oncogene 15, 107-115. MEDLINE

Zhu WY, Jones CS, Kiss A, Matsukuma K, Amin S, De Luca LM. (1997). Exp. Cell. Res. 234, 293-299. MEDLINE

Figure 1 Retinoids inhibit normal breast cell proliferation. Normal breast cells were treated with retinoids in complete medium for 11 h after starving the cells for 24 h. ³H-Thymidine incorporation assay was performed as described in Materials and methods. Each data point was performed in triplet, and the results were reported as c.p.m.±standard error. The single star (*) represents P value less than 0.05, while the double stars (**) represent P value less than 0.01

Figure 2 BMS453 did not cause apoptosis in normal breast cells. (a,b) Normal breast cells were treated retinoids (1 M) for 5 days. TUNEL assay was performed as described in Materials and methods. The numbers of apoptotic cells (darkly stained cells) and the total numbers of cells were counted, and the results expressed as number of apoptotic cell per 1000 cells. (c) Normal breast cells were treated with retinoids for 6 days and an ELISA apoptosis assay was performed as described in Materials and methods. Results are shown as mean of three samples±standard error. The single star (*) represents P value less than 0.05

Figure 3 atRA and BMS453 induce G1 arrest. 184 cells were treated with retinoids (1 M) or vehicle (DMSO) for 5 days and flow cytometry assay was performed as described in Materials and methods

Figure 4 atRA and BMS453 regulate cell cycle modulators. 184 cells were synchronized by depletion of the growth factors and exposed to retinoids (1 M), DMSO, or TGF (2 ng/ml) for the indicated time and Western blot analysis was performed with specific antibodies. Lane 1: 12 h treatment with TGF; lane 2: growing cells; lane 3: starved cells and is also considered as 0 h before treatment; lanes 4-6, 7-9, and 10-12: 24, 48, and 72 h treatment with DMSO, atRA and BMS453

Figure 5 atRA and BMS453 decrease CDK2 kinase activity. 184 cells were synchronized by depletion of the growth factors and exposed to retinoids (1 M), DMSO, or TGF (2 ng/ml) for the indicated time. The cells were lysed and the lysates were immunoprecipitated either with CDK2 antibody and CDK2 kinase assay was performed using histone H1 as a substrate or with cyclin D antibody and cyclin D-associated kinase assay was performed using GST-Rb as a substrate. Lane 1: starved cells and is also considered as 0 h before treatment; lane 2: growing cells; lanes 3-5 and 6-8: 24 and 72 h treatment with DMSO, atRA and BMS453; lane 9: 12 h treatment with TGF

Figure 6 atRA and BMS453 regulate CDK inhibitors. 184 cells were synchronized by depletion of the growth factors and exposed to retinoids (1 M), DMSO, or TGF (2 ng/ml) for the indicated time and Western blot analysis was performed with specific antibodies against CDK inhibitors. Lane 1: 12 h treatment with TGF; lane 2: growing cells; lane 3: starved cells and is also considered as 0 h before treatment; lanes 4-6, 7-9, and 10-12: 24, 48 and 72 h treatment with DMSO, atRA and BMS453

Figure 7 Retinoids regulated CDK2- or CDK4-associated p21 and p15. 184 cells were synchronized by depletion of the growth factors and exposed to retinoids (1 M), DMSO, or TGF (2 ng/ml) for the indicated time. The cells were lysed and the lysates were immunoprecipitated with CDK2 antibody or CDK4 antibody and Western blot analysis was performed. Lanes 1-3 and 4-6: 24 and 72 h treatment with DMSO, atRA and BMS453 and immunoprecipitated with CDK2 or CDK4 antibody

Figure 8 TGF activity in normal breast cells. Normal breast cells were exposed to 1 M atRA or vehicle (DMSO) for 1, 3 or 5 days as indicated. This bioassay for measuring total and active TGF was performed as described in Materials and methods. The data are expressed at relative levels of TGF activity (as reflected by changes in luciferase activity) and are plotted as mean±standard error from at least two independent experiments

Figure 9 TGF-blocking antibody prevents retinoid BMS453-induced inhibition of normal breast cell proliferation. Normal breast cells were first starved of growth factors for 24 h to synchronize the cells. The cells were then treated with retinoids, IgG (40 ng/ml), or an anti-TGF blocking antibody (40 ng/ml) in complete medium for 36 h. ³H-Thymidine incorporation assay was performed as described in Materials and methods. Each data point was performed in triplet, and the results were reported as mean c.p.m.±standard error. Note that the standard error of bars 3 and 8 are so small that the error bars are not seen

Figure 10 Proposed mechanism by which atRA and BMS453 cause G1 arrest