Synergistic Antitumor Activity of Lapatinib and Retinoids on a Novel Subtype of Breast Cancer with Coampliﬁcation of ERBB2 and RARA

Oncogene (2012) 31, 3431–3443 & 2012 Macmillan Publishers Limited All rights reserved 0950-9232/12 www.nature.com/onc ORIGINAL ARTICLE Synergistic antitumor activity of lapatinib and retinoids on a novel subtype of breast cancer with coampliﬁcation of ERBB2 and RARA

G Paroni1, M Fratelli1, G Gardini2, C Bassano2, M Flora2, A Zanetti1, V Guarnaccia1, P Ubezio3, F Centritto1, M Terao1 and E Garattini1

1Laboratory of Molecular Biology, Istituto di Ricerche Farmacologiche ‘Mario Negri’, Milano, Italy; 2Anatomic Pathology Unit, IRCCS Ospedale Santa Maria Nuova, Reggio Emilia, Italy and 3Department of Oncology, Istituto di Ricerche Farmacologiche ‘Mario Negri’, Milano, Italy

All-trans retinoic acid (ATRA), the only clinically Oncogene (2012) 31, 3431–3443; doi:10.1038/onc.2011.506; available cyto-differentiating agent, has potential for the published online 7 November 2011 therapy/chemoprevention of breast carcinoma. Given the heterogeneous nature of this tumor, a rational use of Keywords: targeted therapy; retinoic acid; differentia- ATRA and derivatives (retinoids) in the clinic requires the tion; apoptosis; HER2/neu; lapatinib identification of patients that would benefit from retinoid- based protocols. Here, we demonstrate that 23–32% of the human ERBB2 þ breast cancers show coamplification of retinoic acid receptor alpha (RARA), encoding the Introduction retinoic acid receptor, RARa. This represents a novel subtype of breast cancer characterized by remarkable Breast carcinoma is classified according to the presence/ sensitivity to ATRA and RARa agonists, regardless of absence of estrogen receptor (ER), progesterone recep- positivity to the estrogen receptor, a known modulator of tor and ERBB2 (also known as HER2/Neu) (Rubin and retinoid sensitivity. In estrogen-receptor-negative cellular Yarden, 2001). ERBB2 þ patients used to have a poor models showing coamplification of ERBB2 and RARA, prognosis until the advent of targeted therapy (Baselga, simultaneous targeting of the corresponding gene products 2006; Baselga and Swain, 2009), which relies on with combinations of lapatinib and ATRA causes trastuzumab (Esteva et al.; Mariani et al., 2009) and synergistic growth inhibition, cyto-differentiation and lapatinib (Moreira and Kaklamani, 2010). In spite of apoptosis. This provides proof-of-principle that coampli- this progress, not all patients in this group benefit from fication of ERBB2 and RARA can be exploited for the trastuzumab-based therapies (Hudis, 2007; Esteva et al., stratified and targeted therapy of a novel subtype of breast 2010). In addition, a subset of patients progress after an cancer patients, with an approach characterized by tumor initial response. Innate and acquired resistance to HER2 cell selectivity and low predicted toxicity. The available targeting calls for the development of new and rational cellular models were exploited to define the molecular therapeutic approaches. This is likely to entail the mechanisms underlying the antitumor activity of combina- development of combination strategies involving simul- tions between lapatinib and ATRA. Global gene expres- taneous targeting of determinants crucial for the growth sion and functional approaches provide evidence for three and progression of the tumor (Abramson and Arteaga, components of the antiproliferative/apoptotic responses 2011). triggered by lapatinib þ ATRA. Induction of the retinoid- All-trans retinoic acid (ATRA), the only clinically dependent RARRES3 protein by ATRA stabilizes the useful differentiating agent (Garattini et al., 2007) has effect of lapatinib inhibiting ERBB2 phosphorylation. potential for the therapy/chemoprevention of breast Upregulation and activation of the transcription factor carcinoma (Decensi et al., 2000; Puntoni and Decensi, FOXO3A integrates ATRA-dependent transcriptional 2009). Given the heterogeneous nature of this tumor, a and lapatinib-dependent posttranscriptional signals, con- rational use of ATRA and derivatives (retinoids) in the trolling the levels of effector proteins like the antiapopto- clinic requires the identification of patients that would tic factor, BIRC5. Stimulation of the TGFb pathway by benefit from retinoid-based protocols. ATRA mediates other components of the apoptotic The antitumor activity of retinoids is mediated by process set in motion by simultaneous targeting of ERBB2 specific nuclear receptors (RARa, b, g and RXRa, b, g, and RARa. which are ligand-dependent transcription factors (Kast- ner et al., 1995; Chambon, 1996). In breast cancer cells, RARa expression is controlled by estrogens, and high Correspondence: Dr E Garattini, Laboratory of Molecular Biology, levels of the protein are observed in estrogen-receptor- Istituto di Ricerche Farmacologiche ‘Mario Negri’, via La Masa 19, positive (ER þ ) tumors (Sun et al., 1998). RARa 20156 Milano, Italy. expression determines sensitivity of ER þ and refractori- E-mail: [email protected] À Received 13 August 2011; revised and accepted 1 October 2011; ness of ER breast cancer cells to retinoids (Roman published online 7 November 2011 et al., 1993; Sheikh et al., 1993; Fitzgerald et al., 1997). Antitumor activity of retinoids and lapatinib GParoniet al 3432 Retinoic acid receptor alpha (RARA), the gene causing 50% maximal growth inhibition) (Figure 1c; encoding RARa, maps close (0.65 Mb) to the ERBB2 Supplementary Figure S1). With the exception of BT474, locus, which is generally amplified in ERBB2 þ breast all ER þ cells responded to ATRA in the micromolar carcinoma. Here, we establish that a proportion of range. ERÀ/ERBB2 þ /RARAÀ cells were resistant, ERBB2 þ tumors are characterized by an amplicon whereas the ERÀ/ERBB2 þ /RARA þ cells, SKBR3 and encompassing RARA, which confers retinoid sensitivity AU565, were more sensitive to ATRA (nanomolar upon the ERÀ neoplastic cell. This genetic makeup was range) than the ER þ counterparts. All the cell lines, exploited for concomitant targeting of ERBB2 and regardless of the ER status, synthesized detectable levels RARa, obtaining synergistic antitumor effects. The of RARg and RXRa (Figure 1c), but showed no underlying molecular mechanisms were studied. detectable expression of RARb. ATRA sensitivity was directly correlated to RARa levels in ERÀ cells. The two ERÀ/luminal lines, SKBR3 (ERBB2 þ / Results RARA þ ) and MDA-MB-453 (ERBB2 þ /RARAÀ), were used for subsequent studies (Supplementary Figure A subgroup of breast carcinomas shows ERBB2/RARA S2a). ATRA-dependent growth inhibition was accom- coamplification panied by differentiation of SKBR3 but not MDA-MB- We defined that a proportion of ERBB2 þ breast 453 cells. This effect was detectable at 0.1 mM and carcinomas was characterized by RARA amplification maximal at 1 mM ATRA. Differentiation included (ERBB2 þ /RARA þ ) by real-time quantitative PCR accumulation of oil-red droplets (lactogenesis) (Supple- assays (Q-PCR). Histological sections of 74 tumors mentary Figure S2b) and appearance of adherens positive for ERBB2 by fluorescence in situ hybridiza- junctions containing b-catenin, p120 and VE-cadherin tion (FISH) and 2 ERBB2À controls were analyzed (Supplementary Figure S2c; Dusek and Attardi, 2011). (Figures 1a and b, Supplementary Table S1). Most cases This was accompanied by the induction of the three (76%) proved positive for ERBB2 amplification by Q– proteins (see also Figure 2d) and restructuring of the PCR, when a conservative threshold value was applied. actin cytoskeleton. Growth inhibition and differentia- Coamplification of the RARA locus was found in 23% tion were directly correlated to ligand-dependent activa- of the patients by Q–PCR and 32% by FISH. If samples tion of retinoid receptors, as treatment with ATRA were stratified for ER, no difference in the proportion of caused sustained activation of the transfected retinoid ERBB2 þ /RARA þ patients was evident. reporter, RARE-tk-Luc (Gianni et al., 2009), only in SKBR3 cells (Supplementary Figure S2d). To verify the role of RARA amplification, we compared the activities ERÀ cell lines recapitulating ERBB2 and RARA et al. coamplification are sensitive to retinoids of ATRA (pan-RAR agonist) (Garattini , 2004), the RARa agonist, AM580 (Gianni et al., 1996a), and Five ERÀ and five ER þ breast carcinoma cell lines over- the RARg agonist, CD437 (Gianni et al., 1993), in expressing the ERBB2 protein and two negative controls SKBR3 and MDA-MB-453 cells. AM580 and ATRA (MDA-MB-231 and MCF-7) were evaluated for RARA arrested the growth and induced differentiation of amplification by Q–PCR (Figure 1c). Except for ZR75.1, SKBR3 cells, whereas CD437 was inactive (Supplemen- all ERÀ and ER þ cells with ERBB2 overexpression tary Figures S2a and e). In SKBR3 cells, this is showed amplification of the corresponding gene. The correlated with the activation of transfected RARE-tk- ER þ line, UACC812, and the ERÀ lines, SKBR3 and Luc by the two agonists (Supplementary Figure S2f). AU565, showed ERBB2/RARA coamplification. All ER þ cells contained more RARa than their ERÀ counterparts, and UACC812 synthesized the largest Lapatinib potentiates the activity of ATRA in ERÀ/ amounts of the receptor. In the context of ER- ERBB2 þ /RARA þ cells negativity, the two ERBB2 þ /RARA þ lines, SKBR3 We evaluated whether simultaneous targeting of RARa and AU565, showed significant expression of RARa, and ERBB2 represented a potential therapeutic ap- whereas the ERBB2 þ /RARAÀ lines presented barely proach for ERÀ/ERBB2 þ /RARA þ breast cancer. Lapa- detectable levels of the protein (Figure 1c). tinib, rather than trastuzumab, was chosen as the We evaluated the sensitivity of cell lines to ATRA, ERBB2 inhibitor, as its mechanism of action is defined. À determining their pIC50 (-Log10 of the concentration In the ER cellular context, lapatinib invariably

Figure 1 Retinoic acid receptor alpha (RARA) coamplification in ERBB2 þ breast carcinoma patients and breast carcinoma cell lines. Sensitivity of the cell lines to all-trans retinoic acid (ATRA). (a) Human patients. The copy number of the ERBB2 and RARA genes was determined by quantitative real-time PCR (Q–PCR), using DNA extracted from paraffin-embedded histologic samples. The cutoff used to define amplification of ERBB2 and RARA (3.0) is indicated by horizontal lines. (b) Dual-color fluorescence in situ hybridization (FISH) analysis performed on interphase nuclei. Representative cases with and without ERBB2/RARA coamplification are shown. (c) Cell lines. The two upper graphs show amplification of the ERBB2 and RARA genes in estrogen-receptor-negative (ERÀ) (red columns) and ER þ (green columns) breast cancer cell lines, as determined by Q–PCR. The threshold value for the amplification of the two genes is shown by horizontal lines. Underneath the upper graph the levels of ERBB2 protein, as determined by western blot, are shown. The levels of RARa, RARg, RXRa and b-actin as a loading control are shown below the middle graph. The bottom bar graph shows the sensitivity of each cell line to the antiproliferative activity of ATRA. Data are expressed in –Log10 of the IC50 values calculated (pIC50).

Oncogene Antitumor activity of retinoids and lapatinib G Paroni et al 3433

Oncogene Antitumor activity of retinoids and lapatinib GParoniet al 3434 inhibited ERBB2 þ cell growth (Figure 2a). Consistent with retinoid sensitivity and the relative levels of RARa, ATRA potentiated lapatinib activity only in ERBB2 þ / RARA þ cells (SKBR3 and AU565). In SKBR3 cells, the antiproliferative effect of lapatinib and ATRA was synergistic (Figure 2b) and lapatinib þ ATRA coopera- tion extended to lactogenic (Figure 2c) and epithelial differentiation (Figure 2d). In the ER þ context, ATRA enhanced growth inhibition by lapatinib in the retinoid- sensitive ERBB2 þ cell lines, regardless of RARA amplification (see UACC812 as an example in Figure 2a). In SKBR3 cells, lapatinib þ ATRA induced cytotoxicity, which was not observed with lapatinib or ATRA alone (see sub-G1 fraction in Figure 3a). This was the result of apoptosis, as indicated by PARP/caspase-3 cleavage (Figure 3b). Apoptosis was confirmed by a specific assay (Paroni and Brancolini, 2011) based on overexpression of a mitochondria-targeted GFP protein containing caspase-dependent cleavage sites (DD-GFP) or inactive mutants, thereof (AA-GFP) serving as negative controls (Figure 3c). RARA amplification was responsible for the cytotoxic/apoptotic response of SKBR3 cells, as cytotoxicity was replicated only by lapatinib þ AM580 (Figure 3d). More importantly, RARa silencing reduced the DD-GFP cleavage observed with lapatinib þ ATRA (Figure 3e).

ATRA and lapatinib cross talk perturbs the SKBR3 cell transcriptome We determined the transcriptomes of SKBR3 cells treated for 12 and 48 h with lapatinib, ATRA or the combination. A set of 2,846 genes (active dataset) was modiﬁed by one of the factors (lapatinib, ATRA or interaction; two-way ANOVA, Po10À5) in at least one of the time points considered (Supplementary Table S2). Principal component analysis indicated that the transcriptional effect of lapatinib was short-lived, with

Figure 2 Targeting of ERBB2 with lapatinib potentiates the activity of all-trans retinoic acid (ATRA) in ERBB2 þ /RARA þ breast carcinoma cells. (a) Cell lines were treated with vehicle (DMSO), lapatinib (0. 1 or 1.0 mM), ATRA (0. 1 or 1.0 mM) or the combination. The number of viable cells was determined after trypan blue staining. Each value is the mean±s.d. of three culture dishes. **Signiﬁcantly different relative to lapatinib and ATRA treatments (Student’s t-test, Po0.001). Red ¼ ERÀ cells; green- ¼ ER þ cells. (b) SKBR3 cells were treated with vehicle (DMSO), lapatinib, ATRA or lapatinib þ ATRA for 96 h and subjected to the sulforhodamine assay. The two graphs show the isobolograms of the combination at the IC50 and IC70 levels. The additivity lines separate the antagonistic (upper) from the synergistic (lower) region. (c) SKBR3 cells were treated as in (a) for 96 h. The picture illustrates Oil-red O accumulation in cells. The graph shows the quantitative results obtained after extraction of Oil-red O. Results are expressed in fold induction relative to the control value taken as 1 and are the mean±s.d. of three culture dishes. **Signiﬁcantly different (Po0.01, Student’s t-test). (d) SKBR3 cells were treated as in (a) for 48 h. Left: cells were evaluated for the appearance of adherens junctions using anti-VE-cadherin antibodies. Right: expression of the adherens junction associated proteins, VE-cadherin, b-catenin and p-120, was assessed by western blot. Each result is representative of two experiments.

Oncogene Antitumor activity of retinoids and lapatinib G Paroni et al 3435

Figure 3 Cotreatment with lapatinib and all-trans retinoic acid (ATRA) induces apoptosis by RARa.(a) SKBR3 cells were treated with vehicle (DMSO), lapatinib (0.1 mM), ATRA (0.1 mM) or the combination. Upper: representative FACS analyses after 72 h. Lower: time course for the appearance of sub-G1 cells. (b) Cells were treated as in (a). Cleavage of PARP and caspase-3 was determined by western blot. (c) Cells expressing DD-GFP or the AA-GFP negative control were treated as in (a) for 48 h. DD-GFP or AA-GFP cleavage was assessed by western blot. Upper: the panel shows a representative western blot. Lower: apoptosis was evaluated by counting GFP-positive cells showing a diffused GFP signal. Each value is the mean±s.d. of three cultures. (d) Cells were challenged with AM580 (0.1 mM), CD437 (0.1 mM) and ATRA (0.1 mM) alone or in combination with lapatinib (0.1 mM) for 72 h. Cell growth and viability were evaluated after trypan blue staining. Each result is the mean±s.d. of three cultures. **Significantly different from ATRA and lapatinib (Student’s t-test; Po0.01). (e) Cells were cotransfected with a RARa-targeting small interfering RNA (siRNA) or the negative control (NC) and DD-GFP. Cells were treated as in (a). RARa silencing was confirmed 24 h after transfection by western blot (left panel). DD-GFP cleavage was measured by western blot (middle panel) and quantified by densitometry (right panel). The results are expressed as the DD-GFP/DDD-GFP signals after normalization. The results are representative of three experiments. major perturbations observed after 12 h and reversion to extracting four clusters according to the expression baseline at 48 h (Supplementary Figure S3a). ATRA profile (Supplementary Figure S3b). Cluster 1, compris- exerted a prolonged action regardless of lapatinib ing genes regulated by lapatinib but not by ATRA, was addition. ANOVA was followed by time-course analysis enriched in genes controlling cell cycle and DNA considering all the possible contrasts (Baesian Estima- replication (Supplementary Figure S3c). The overall tion of Temporal Regulation) (Aryee et al., 2009). We effect of lapatinib was repressive, causing downregula- focused on the subgroup of genes modulated by tion of 75% of the modulated genes. In the presence of lapatinib þ ATRA relative to lapatinib or ATRA alone, lapatinib alone, most genes modified at 12 h returned to

Oncogene Antitumor activity of retinoids and lapatinib GParoniet al 3436 baseline or showed opposite regulation at 48 h, whereas Short-term treatment (1 h) with lapatinib suppressed induction or repression was maintained with lapati- phosphorylation/activation of ERBB2, resulting in nib þ ATRA. This was consistent with principal com- hypo-phosphorylation of the downstream kinases, ponent analysis and indicated that ATRA stabilized the AKT and ERK2 (Figure 4b, left). These effects were transcriptional effects of lapatinib. Cluster 2 contained not altered by ATRA. Inhibition of ERBB2, AKT and genes regulated by ATRA but not by lapatinib ERK2 phosphorylation by lapatinib was short-lived (Supplementary Figure S3d). The combination exerted reverting to baseline by 48 h (Figure 4b, right). In a potentiating effect on ATRA-dependent expression of contrast, ATRA caused a late inhibitory effect on many genes. The type of genes enriched in this cluster ERBB2 and downstream kinases from 24 h onwards. was heterogeneous, although controllers of development In the presence of lapatinib, this resulted in long-lasting and cell growth stood out. Cluster 3 consisted of genes suppression of ERBB2 phosphorylation. modulated by lapatinib and ATRA concomitantly Among the retinoid targets in the active dataset, (Supplementary Figure S3e). The majority was upregu- retinoic acid receptor responder 3 (RARRES3) (Huang lated by ATRA and cotreatment with lapatinib resulted et al., 2000; Sturniolo et al., 2003; Shyu et al., 2005; Tsai in additive/synergistic effects. Among the enriched et al., 2007, 2009; Barral et al., 2009) was one of the genes, those involved in lipid metabolism and apoptosis genes with the highest induction ratio (combination vs were relevant. Cluster 4 gathered genes activated or control at 48 h). The effect of the combination was repressed only by lapatinib þ ATRA, and was enriched significantly different from single drug treatment for determinants of cell cycle and apoptosis (Supple- (Figure 4c, Supplementary Table S2). Considering that mentary Figure S3f). RARRES controls the levels of ERBB2 in ovarian cancer (Ou et al., 2008), the corresponding gene was silenced to evaluate its significance for the long-term Inhibition of ERBB2 by lapatinib enhances the regulation of ERBB2 by ATRA in SKBR3 cells. transcriptional responses to ATRA Silencing of RARRES3 reduced the delayed ATRA- We studied the molecular mechanisms underlying the dependent inhibition of ERBB2 phosphorylation action of lapatinib on the retinoid-dependent arm of the (Figure 4d). This effect was not the consequence of a cross talk. In SKBR3 cells, ATRA reduced the amounts change in the amounts of the protein. of the RARa protein (Supplementary Figure S4a), without affecting the corresponding transcripts (Supple- mentary Figure S4b). This was consistent with activa- TGFb, FOXO3A and SMAD3 are central to the tion and subsequent proteasome-dependent degradation synergism between lapatinib and ATRA of the receptor (Gianni et al., 2009). Lapatinib too To further identify molecular determinants of the cross downregulated the RARa protein, resulting in additive talk between lapatinib and ATRA, we performed effects upon cotreatment with the retinoid. These effects focused network analysis on the active dataset.We suggested increased activation of RARa by lapatinib. In linked prominent breast carcinoma growth factors to the same conditions, modulation of RARg by lapatinib transcription factors, and downstream gene products and/or ATRA was not observed. Consistent with this involved in cell cycle and apoptosis. The network was idea, the expression profiles of retinoid target genes subsequently integrated with the RARa, ERBB2, ERK2 (Balmer and Blomhoff, 2002, 2005), many of which were À5 and AKT nodes (Supplementary Figure S5). indeed regulated by ATRA in SKBR3 cells (Po10 , TGFb2 is a growth factor and mediates the action of ANOVA), showed potentiated expression upon addition ATRA in various cell types (Zujewski et al., 2001; of lapatinib (Supplementary Figure S4c). However, Dokmanovic et al., 2002; Danforth, 2004; Mucida and direct stimulation of RARa by lapatinib was unlikely Cheroutre, 2007; Wang et al., 2007; Xiao et al., 2008). In to explain the observed effects. In fact, although ATRA SKBR3 cells, TGFb2 and the antimetastatic receptor, and the specific RARa agonist, AM580, activated the TGFBRII (Yang et al., 2008b; Tan et al., 2009), were retinoid-dependent RARE-tk-Luc reporter in SKBR3 upregulated by ATRA and were further induced by cells, activation was not further enhanced by the lapatinib þ ATRA (Supplementary Figure S5). This addition of lapatinib (Supplementary Figure S4d). implied that some of the pathways activated by lapatinib Similar results were obtained in MDA-MB-453 cells and/or ATRA involved TGFb mediation. overexpressing RARa and RARE-tk-Luc. SMAD family member 3 (SMAD3) and forkhead box O3 (FOXO3A) were among the top 10 transcription Control of ERBB2 phosphorylation by RARRES3 is a factors regulated by ATRA and/or lapatinib (Supple- component of the cross talk between lapatinib and ATRA mentary Figure S5 and Supplementary Table S2), for Temporal regulation of cell cycle genes suggested that which there was significant overconnection in the active ATRA prolonged the antiproliferative effect of lapatinib dataset (Supplementary Table S3). The two groups of and this was confirmed experimentally (Figure 4a). In FOXO3A and SMAD3 target genes were characterized SKBR3 cells, lapatinib induced a transient increase (24– by the presence of five common elements involved 48 hours) of the G1-fraction at the expenses of the S- in cell growth and apoptosis (CDKN2B, CDKN2D, fraction, whereas ATRA exerted a late G1-arrest. The FOXM1, BCL2L11 and BIRC5/survivin) (Supplemen- combination enhanced and stabilized the lapatinib- tary Figures S6a and b). The expression profile of dependent G1-block. these genes was consistent with a role in integrating

Oncogene Antitumor activity of retinoids and lapatinib G Paroni et al 3437

Figure 4 Cross talk between lapatinib and all-trans retinoic acid (ATRA) in SKBR3 cells: the lapatinib arm. (a) Cells were treated with vehicle (DMSO), lapatinib (0.1 mM), ATRA (0.1 mM) or the combination for the indicated amount of time. From left to right: the number of viable cells was determined. Each value is the mean±s.d. of three culture dishes. Cells were also subjected to FACS analysis. The proportion of cells in the various phases of the cell cycle (G1, S and G2/M) is indicated. (b) Cells were treated as in (a) for 1 h (left) or for the indicated amount of time (right). The levels and the phosphorylation/activation state of ERBB2, AKT and ERK2 were determined by western blot analysis. Actin is shown as a loading control. The results shown are representative of two independent experiments. (c) Cells were treated as in (a) for 12 and 48 h. Expression of the RARRES3 mRNA was determined by Q–PCR analysis. Each value is the mean±s.d. of three replicate cultures. (d) Cells were transfected with small interfering RNAs (siRNAs) targeting RARRES3 or the appropriate negative control siRNAs (NC). After 24 h, cells were treated as in (a) for a further 32 h. Left: specific silencing of RARRES3 was evaluated by Q–PCR as in (c) in vehicle treated cells 24 h after transfection. Each value is the mean±s.d. of three replicate cultures. Right: determination of activated phospho-ERBB2 and total ERBB2 by western blot. Tubulin is shown as a loading control. Each result is representative of two independent experiments. growth inhibitory and apoptotic signals originating induction of SMAD3 and FOXO3A was dependent on from SMAD3 and FOXO3A. RARa activation. In SKBR3 cells, SMAD3 or FOXO3A mRNAs and proteins were upregulated predominantly by ATRA, although further upregulation was afforded by lapati- SMAD3 activation is TGFb-dependent, whereas nib þ ATRA, confirming and extending the microarray activation of FOXO3A integrates transcriptional results (Figures 5a and b). Induction of SMAD3 and and posttranscriptional signals FOXO3A was also observed in the other ERÀ/ERBB2 þ / We evaluated whether lapatinib þ ATRA affected the RARA þ line, AU565, but not in the ERÀ/ERBB2 þ / biological activity of SMAD3 and FOXO3A (Figure 5c RARAÀ counterparts, MDA-MB453 and HCC1569 and d). Treatment of SKBR3 cells with ATRA caused (Supplementary Figure S6c), strengthening their rele- phosphorylation/activation of SMAD3, which paral- vance for the cross talk between lapatinib and ATRA. leled induction of the protein. Phosphorylation of FOXO3A and SMAD3 proteins were induced by SMAD3 was left unaffected by lapatinib, regardless of AM580, whereas neither CD437 nor the RARb agonist, ATRA. Induction/activation of SMAD3 by ATRA may BMS641, exerted any significant effect (Supplemen- be indirect and linked to TGFb induction, as SMAD3 tary Figure S6d). Additionally, silencing of RARa is a TGF signal-transducer (McEarchern et al., 2001; blunted the induction of both SMAD3 and FOXO3A Buck and Knabbe, 2006). Blockade of the TGF recep- (Supplementary Figure S6e). Thus, ATRA-regulated tors with SB431542 did not alter the induction of

Oncogene Antitumor activity of retinoids and lapatinib GParoniet al 3438 SMAD3 (Figure 5c). In contrast, SB431542 blunted the nucleus (Figure 5d) without affecting FOXO3A protein phosphorylation/activation of SMAD3 observed with expression levels (Figure 5b). In contrast, ATRA did ATRA and lapatinib þ ATRA. Thus, ATRA was not alter the intracellular compartmentalization of necessary and sufﬁcient to cause both activation and FOXO3A (Figure 5d), although it induced the protein induction of the SMAD3 protein, but only activation signiﬁcantly (Figure 5b). Lapatinib-dependent FOX- required TGFb. O3A sequestration into the nucleus was a rapid process FOXO3A activation was assessed by measuring its completed within 1 h (Supplementary Figures S7a and subcellular localization (Afonja et al., 2004). Lapatinib b). Thus, the result of lapatinib þ ATRA treatment caused complete relocalization of FOXO3A into the was the sum of the two effects exerted separately by ATRA and lapatinib, that is, accumulation of large amounts of FOXO3A and retention into the nucleus. The action of lapatinib likely involved inhibition of AKT-dependent FOXO3A phosphorylation, with con- sequent sequestration into the nucleus and protection from proteasome-dependent degradation (Yang and Hung, 2009). In conclusion, FOXO3A represents a central effector of the cross talk, integrating transcriptional induction by ATRA with posttranslational activation by lapatinib.

TGF, RARRES and FOXO3A contribute to the apoptotic response activated by lapatinib þ ATRA We evaluated the role of TGFb, FOXO3A and RARRES3 in the apoptotic response triggered by lapatinib þ ATRA. The TGF-receptor inhibitor, SB431542, protected SKBR3 cells from the apoptotic response triggered by lapatinib þ ATRA, as assessed by the DD-GFP cleavage assay (Figure 6a, left). Protection was likely to be the consequence of a predominant action on the retinoid arm of the cross talk, as suggested by the effects of SB431542 on the antiproliferative action of lapatinib, ATRA and the combination (Figure 6a, right). SB431542 did not alter lapatinib-dependent growth inhibition, whereas it reduced the antiproliferative effect of ATRA. This resulted in partial reversion of the cumulative action of lapatinib þ ATRA. Silencing experiments demonstrated that FOXO3A contributed to the apoptotic process activated by combined targeting of ERBB2 and RARa in SKBR3 cells (Figure 6b). Part of this effect involved regulation of effector proteins like BIRC5. BIRC5 is a FOXO3A target gene coding for an antiapoptotic factor (Yang et al., 2008a; Guha et al., 2009). Downregulation of BIRC5 mRNA

Figure 5 SMAD3/FOXO3A induction and activation by lapatinib and all-trans retinoic acid (ATRA). (a) and (b) SKBR3 cells were treated with vehicle (DMSO), lapatinib (0.1 mM), ATRA (0.1 mM) or the combination. The graphs show the levels of SMAD3 and FOXO3A mRNAs determined by Q–PCR. Each value is the mean±s.d. of three replicates. The western blot illustrates the amounts of FOXO3A and SMAD3 proteins determined after 48 h. Tubulin serves as a loading control. (c) SKBR3 cells were treated as in (a) for 18 h, after preincubation in the presence or absence of SB431542 (10 mM) for 1 h. The amounts of FOXO3A/SMAD3 proteins and SMAD3 phosphorylation were assessed by western blot. Tubulin was used as a loading control. The results are representative of two experiments. (d) Immuno-ﬂuorescence localization of the FOXO3A protein in cells treated as in (a) for 24 h. The results are representative of two experiments. Left ¼ anti- FOXO3A antibody; Middle ¼ Hoechst; Right ¼ merge.

Oncogene Antitumor activity of retinoids and lapatinib G Paroni et al 3439

Figure 6 TGF (transforming growth factor), FOXO3A and RARRES3 contribute to apoptosis triggered by lapatinib þ ATRA. (a) Left: SKBR3 cells expressing DD-GFP were treated with vehicle or the combination of lapatinib (0.1 mM) and all-trans retinoic acid (ATRA) (0.1 mM) in the absence/presence of SB431542 (10 mM), for 48 h. DD-GFP and PARP cleavage (DPARP) was measured by western blot. Tubulin was used as a loading control. The results are representative of two experiments. Right: cell growth and viability were determined with sulforhodamine. Each value is the mean±s.d. of six cultures, and the data are representative of two experiments. **Signiﬁcantly different (Po0.01, Student’s t-test) from controls without SB431542. (b, e) Left: cells were cotransfected with DD-GFP and small interfering RNAs (siRNAs) targeting RARRES3, FOXO3A or negative controls (NC) before treatment with vehicle (V), lapatinib (0.1 mM) and/or ATRA (0.1 mM). FOXO3A and RARRES3 mRNAs were determined by Q–PCR. The results are the mean±s.d. of three replicates. Middle-right: DD-GFP cleavage was measured by western blot and quantiﬁed by densitometry. The results are expressed as the DD-GFP/DDD-GFP signals after normalization [100% is the signal observed in NC transfected cells treated with vehicle (V)] (c) BIRC5 protein was determined by western blot after treatment of cells as in (a). Actin was used as a loading control. (d) Cells were transfected with FOXO3A-targeting siRNAs and the negative control (NC) before treatment with lapatinib and/ or ATRA as in (a) for 48 h. Left: FOXO3A mRNA was determined by Q–PCR. Right: BIRC5 was determined by western blot. Tubulin was used as a loading control.

(Supplementary Figure S5) and protein (Figure 6c) by crucial determinant in the apoptotic process activated by lapatinib and ATRA was enhanced and prolonged by the combinations of lapatinib and ATRA. combination. Silencing of FOXO3A attenuated downregulation of BIRC5 by lapatinib and/or ATRA (Figure 6d). In a similar fashion, silencing of RARRES3 was accompanied by protracted inhibition of the DD-GFP Discussion cleavage observed with lapatinib þ ATRA (Figure 6e). This indicated that long-term inhibition of ERBB2 by Retinoids have been proposed in the adjuvant treatment ATRA via RARRES3 induction is likely to represent a of breast carcinoma for their ability to inhibit growth

Oncogene Antitumor activity of retinoids and lapatinib GParoniet al 3440 and induce morphological or phenotypic differentiation predominantly involved in mediating some of the of breast carcinoma cell lines (Yang et al., 2002; Paik cellular responses activated by ATRA. Interestingly, et al., 2003). An effective clinical use of retinoids in retinoid-dependent induction and activation of the breast carcinoma requires the identification of sensitive signal transducer, SMAD3, is not necessary for the patient subpopulations, as this tumor is highly hetero- apoptotic component of TGFb activity in our cellular geneous. The goal can be achieved in a rational way by system, as silencing of the gene did not alter DD-GFP definition of the molecular determinants controlling the cleavage triggered by lapatinib þ ATRA (data not sensitivity of the neoplastic cell to ATRA and deriva- shown). As for FOXO3A, full activation of the tives. Preclinical and clinical data indicate that the levels transcription factor involves direct induction by ATRA of RARa may represent one such determinant, as and sequestration into the nucleus by lapatinib. In our retinoid-responsive ERa þ breast carcinoma cells con- model (Supplementary Figure S8), these combined tain much higher levels of the receptor than retinoid- effects result in the regulation of various effector refractory ERaÀ cells (Roman et al., 1993; Sheikh et al., molecules potentially relevant for the growth inhibitory 1993; Fitzgerald et al., 1997). and apoptotic effects of the combination between In this report, we establish that a significant fraction lapatinib and ATRA. On the basis of our results, we of all ERBB2 breast carcinomas is characterized by propose that BIRC5 downregulation is an effector of coamplification of the RARA locus. Amplification of FOXO3A-dependent apoptosis of SKBR3 cells. RARA leads to increased expression of an active form of RARRES3, TGFb and FOXO3A define three path- the corresponding RARa protein, and is associated with ways contributing to the apoptotic program triggered by sensitivity to the antiproliferative/cyto-differentiating lapatinib þ ATRA in ERÀ/ERBB2 þ /RARA þ cells. The action of ATRA. This is of particular relevance in the level of integration among these pathways is incomple- context of ERaÀ tumors, which are refractory to tely understood, although experiments, which were not retinoids. In ERÀ/ERBB2 þ /RARA þ cells, sensitivity to the object of this work demonstrated that RARRES3 the retinoid is further enhanced by the tyrosine kinase silencing did not alter FOXO3A mRNAs expression/ inhibitor, lapatinib, which targets ERBB2 and EGFR1 regulation. Conversely, silencing of FOXO3A had no (epidermal growth factor receptor 1). In our experi- effect on RARRES3 mRNA expression, in line with mental conditions, the enhancing effect of lapatinib is direct transcriptional activation of the RARRES3 gene likely to be the result of a specific action on ERBB2, as by ATRA and lapatinib þ ATRA. synergistic inhibition of SKBR3 cell growth was recently In conclusion, our work provides the initial rationale reported also for combinations of the specific ERBB2 for the design of clinical studies aimed at evaluating the inhibitor, trastuzumab and ATRA (Koay et al, 2010). feasibility of our approach based on simultaneous Our work represents proof-of-concept that simultaneous targeting of ERBB2 and RARa for the stratified therapy targeting of ERBB2 and RARa has the potential to be of ERÀ/ERBB2 þ /RARA þ breast cancer patients. It also exploited in the treatment of a subpopulation of breast provides new insights into the biology of this subtype of carcinoma patients characterized by coamplification of tumor, defining new molecular determinants and in- the corresponding two genes. In principle, combinations tracellular pathways involved in the proliferation/ between ATRA or RARa agonists and lapatinib are survival of the neoplastic cell. promising for at least two reasons. First, combined targeting of ERBB2 and RARa results in highly synergistic interactions. Second, treatment with the combination is likely to target the cancer cell selectively. Materials and methods These two actions are foreseen to increase the therapeutic index of the retinoid. Current experimental work ERBB2 and RARA copy number assessment by FISH aims at strengthening the rationale of our therapeutic and Q–PCR analyses approach in vivo. However, this will require the FISH assays of ERBB2 were performed on 4-mm paraffin generation of novel preclinical models recapitulating sections using the PathVysion kit (Abbott Laboratories, Abbott Park, IL, USA). The ERBB2 probe was labeled with the coamplification of ERBB2 and RARA, as the cell SpectrumOrange, and the CEP17 control probe (D17Z1, lines currently available do not give rise to tumor 17p11.1-q11.1) with SpectrumGreen. In the case of FISH xenografts in immune-deficient animals. assays on RARA, paraffin sections were hybridized with the In ERBB2 þ /RARA þ cells, the cross talk between Smith-Magenis assay probe set, which includes a 17p11.2 lapatinib and ATRA is two-way. The action exerted by DNA probe labeled with SpectrumOrange, encompassing the ATRA on the ERBB2 arm of the cross talk involves SHMT1, TOP3, FLII and LLGL1 loci, and a 17q21.2 DNA sustained and prolonged inhibition of ERBB2 phos- probe labeled with SpectrumGreen, encompassing the GRB7, phorylation/activation by induction of RARRES3. MLN51, SHGC-146999, THRA and RARA loci (Abbott Lapatinib affects the retinoid-dependent arm of the Laboratories). cross talk in a more complex manner and direct DNA from cell lines was prepared with the DNeasy Blood & Tissue Kit (Qiagen, Valencia, CA, USA). DNA from paraffin- modulation of RARa transcriptional activity by lapati- embedded patient samples was obtained using the QuickEx- nib by the inhibition of ERBB2-regulated AKT is tract FFPE DNA Extraction Kit (Epicentre Biotechnologies, unlikely (Srinivas et al., 2006). Madison, WI, USA). ERBB2 and RARA copy number was We identified not only RARRES3 but also TGFb and quantified by real-time quantitative PCR (Q–PCR), using the FOXO3A as crucial players in the cross talk. TGFb is 7300 (SDS software v 1.3) System (Applied Biosystems, Foster

Oncogene Antitumor activity of retinoids and lapatinib G Paroni et al 3441 City, CA, USA). Predesigned TaqMan Copy Number Assays Small interfering RNA experiments Hs01079964_cn (RARA) and Hs05520410_cn (ERBB2) were The following stealth RNAi (Invitrogen) were used: RARa from Applied Biosystems. The ribonuclease P RNA compo- RNAi (HSS109061); FOXO3A RNAi (VHS41092); RARRES3 nent H1 gene (RPPH1), located on chromosome 17q11.2 was RNAi (HSS109078) and SMAD3 RNAi (HSS106252). Stealth used as reference control. Samples were analyzed in triplicate. RNAi Negative Control duplexes (medium GC, cat. no. Human genomic DNA (Cat.G1471) (Promega, Madison, WI, 12935-300 and high GC, cat. no. 12935-400) were used as USA) was run in every assay as a calibrator sample. Further negative controls. Cells were transfected with stealth RNAi details are available in Supplementary Methods. oligonucleotides (60 nM) in OptiMem medium containing Lipofectamine 2000 (Invitrogen). Cell cultures, chemicals and fluorescence-activated cell sorting analysis Reporter gene assays and DD-GFP apoptotic assay BT474 and ZR75.1 were a kind gift from Dr Elda Tagliabue The RARE-containing reporter gene (RARE-tk-Luc) and the (Istituto Nazionale dei Tumori, Milano, Italy), while EFM- RARa expressing plasmids have already been described (Gianni 192A cells were purchased from DSMZ (Deutsche Sammlung et al., 1996a, 2009). For the luciferase assays, cells were trans- von Mikroorganismen und Zellkulturen GmbH, Braunschweig, fected at 30–40% confluence with the indicated mammalian Germany). All the other cell lines were obtained from the expression plasmids. Cell lysates were prepared 24 h after ATCC (American Type Culture Collection, Manassas, VA, transfection, and luciferase activity was measured. Values were USA). Except for HCC1569 (basal A) and MDA-MB231 (Basal always normalized for the Renilla luciferase activity using the B), all the cell lines are characterized by a luminal phenotype dual-luciferase reporter assay system (Promega). The apoptotic (35). SKBR3 cells stably transfected with the DD-GFP plasmid assay based on DD-GFP cleavage (Paroni and Brancolini) was were selected and maintained in 250 mg/ml G418 (Invitrogen, performedasdetailedinSupplementaryMethods. Carlsbad, CA, USA). Cell cultures were grown in phenol red- free Dulbecco’s modified Eagle’s medium supplemented with Gene expression microarray assays and Q–PCR 5% fetal calf serum, glutamine (2 mM)at371Cin5%CO2 Gene-expression microarray results were deposited in the atmosphere. ATRA was from Sigma (ST Louis, MI, USA) and Array Express database (ArrayExpress accession: E-MEXP- lapatinib from LC Laboratories (Woburn, MA, USA). Viability 3192, http://www.ebi.ac.uk/arrayexpress/). Further details on was measured by the trypan blue exclusion and the sulforho- the experimental procedures and post-hoc analysis of the data damine B assays. The RAR specific agonists, CD437 and are available in the Array Express database and in Supple- AM580, were previously described (Gianni et al., 1996b), and mentary Methods. Selected genes were found to be differen- the TGF-receptor kinase inhibitor, SB-431542, was from Sigma. tially expressed, and were validated by Q–PCR using the 7300 Combination effects of lapatinib and ATRA were measured system (Applied Biosystems) and TaqMan technology. The with the isobologram method (Colombo et al., 2011). Flow- Taqman assays performed included FGFR3, Hs00179829_m1; cytometric (fluorescence-activated cell sorting) analysis of DNA IGFBP3, Hs00181211_m1; BMF, Hs00372937_m1; FOXO3A, content was performed as described (Lupi et al., 2004). Hs00818121_m1; RARRES3, Hs00184937_m1; SMAD3, Hs0096920_m1; and 18S, Hs99999901_s1. Data were analyzed by comparative threshold cycle approaches using the 18S as a Western blots control. Western blots were performed as detailed (Terao et al, 2011) with the antibodies described in Supplementary methods. Blots were developed and analyzed using an automated fluorescence scanner (Typhoon, GE Healthcare, Fairfield, CT, USA). Conflict of interest

The authors declare no conflict of interest. Immunofluorescence, time-lapse analysis and oil red O assay For indirect immunofluorescence microscopy, cells were fixed with 3% paraformaldehyde and permeabilized with 0.1% Triton X 100 in phosphate-buffered saline for 5 min. Fixed Acknowledgements cells were incubated for 45 min with the anti-VE-cadherin, anti-p120, anti-b-catenin, and anti-FOXO3A antibodies. The experimental work of our talented PhD student, Roberta Positive cells were identified following further incubation for Affatato, is acknowledged. We would like to thank Silvio 30 min with 488 or 543 Alexa-conjugated secondary antibodies Garattini and Rony Seger (Weizmann Institute, Rehovot, (Invitrogen). Time-lapse analysis of FOXO3A-GFP or DD- Israel) for the critical reading of the manuscript. Grants from GFP transfected SKBR3 cells was performed using the the Associazione Italiana per la Ricerca contro il Cancro Imaging Station cell^R (Olympus, Center Valley, PA, USA). (AIRC), the Fondazione Italo Monzino and the Negri– The FOXO3A-GFP plasmid was kindly provided by Mien- Weizmann Foundation were fundamental for the completion Chie Hung (The University of Texas). Oil red O assays were of this work. The work was also partially supported with performed as described by Mueller et al. (1998). grants from the Ministero della Salute.

References

Abramson V, Arteaga CL. (2011). New strategies in HER2-over- by RAR agonists, antiestrogen and HER-2/neu antagonist in expressing breast cancer: many combinations of targeted drugs breast cancer cells. Evidence for a role in apoptosis. Oncogene 23: available. Clin Cancer Res 17: 952–958. 8135–8145. Afonja O, Juste D, Das S, Matsuhashi S, Samuels HH. Aryee MJ, Gutierrez-Pabello JA, Kramnik I, Maiti T, Quackenbush J. (2004). Induction of PDCD4 tumor suppressor gene expression (2009). An improved empirical bayes approach to estimating differential

Oncogene Antitumor activity of retinoids and lapatinib GParoniet al 3442 gene expression in microarray time-course data: BETR (Bayesian Guha M, Plescia J, Leav I, Li J, Languino LR, Altieri DC. (2009). Estimation of Temporal Regulation). BMC Bioinformatics 10: 409. Endogenous tumor suppression mediated by PTEN involves Balmer JE, Blomhoff R. (2002). Gene expression regulation by retinoic survivin gene silencing. Cancer Res 69: 4954–4958. acid. J Lipid Res 43: 1773–1808. Huang SL, Shyu RY, Yeh MY, Jiang SY. (2000). Cloning and Balmer JE, Blomhoff R. (2005). A robust characterization of retinoic characterization of a novel retinoid-inducible gene 1(RIG1) deriving acid response elements based on a comparison of sites in three from human gastric cancer cells. Mol Cell Endocrinol 159: 15–24. species. J Steroid Biochem Mol Biol 96: 347–354. Hudis CA. (2007). Trastuzumab—mechanism of action and use in Barral PM, Sarkar D, Su ZZ, Barber GN, DeSalle R, Racaniello VR clinical practice. N Engl J Med 357: 39–51. et al. (2009). Functions of the cytoplasmic RNA sensors RIG-I and Kastner P, Mark M, Chambon P. (1995). Nonsteroid nuclear MDA-5: key regulators of innate immunity. Pharmacol Ther 124: receptors: what are genetic studies telling us about their role in 219–234. real life? Cell 83: 859–869. Baselga J. (2006). Targeting tyrosine kinases in cancer: the second Koay DC, Zerillo C, Narayan M, Harris LN, DiGiovanna MP. (2010). wave. Science 312: 1175–1178. Anti-tumor effects of retinoids combined with trastuzumab or Baselga J, Swain SM. (2009). Novel anticancer targets: revisiting tamoxifen in breast cancer cells: induction of apoptosis by retinoid/ ERBB2 and discovering ERBB3. Nat Rev Cancer 9: 463–475. trastuzumab combinations. Breast Cancer Res 12: R62. Buck MB, Knabbe C. (2006). TGF-beta signaling in breast cancer. Lupi M, Matera G, Branduardi D, D0Incalci M, Ubezio P. (2004). Ann NY Acad Sci 1089: 119–126. Cytostatic and cytotoxic effects of topotecan decoded by a novel Chambon P. (1996). A decade of molecular biology of retinoic acid mathematical simulation approach. Cancer Res 64: 2825–2832. receptors. FASEB J 10: 940–954. Mariani G, Fasolo A, De Benedictis E, Gianni L. (2009). Trastuzumab Colombo V, Lupi M, Falcetta F, Forestieri D, D0Incalci M, as adjuvant systemic therapy for HER2-positive breast cancer. Nat Ubezio P. (2011). Chemotherapeutic activity of silymarin Clin Pract Oncol 6: 93–104. combined with doxorubicin or paclitaxel in sensitive and multi- McEarchern JA, Kobie JJ, Mack V, Wu RS, Meade-Tollin L, Arteaga drug-resistant colon cancer cells. Cancer Chemother Pharmacol CL et al. (2001). Invasion and metastasis of a mammary tumor 67: 369–379. involves TGF-beta signaling. Int J Cancer 91: 76–82. Danforth Jr DN. (2004). All trans-retinoic acid acts synergistically Moreira C, Kaklamani V. (2010). Lapatinib and breast cancer: current with hydroxytamoxifen and transforming-growth factor beta to indications and outlook for the future. Expert Rev Anticancer Ther stimulate apoptosis in MCF-7 breast cancer cells. J Endocrinol 183: 10: 1171–1182. 395–404. Mucida D, Cheroutre H. (2007). TGFbeta and retinoic acid intersect Decensi A, Bonanni B, Guerrieri-Gonzaga A, Torrisi R, Manetti L, in immune-regulation. Cell Adhes Migr 1: 142–144. Robertson C et al. (2000). Chemoprevention of breast cancer: the Mueller E, Sarraf P, Tontonoz P, Evans RM, Martin KJ, Zhang M Italian experience. J Cell Biochem Suppl 34: 84–96. et al. (1998). Terminal differentiation of human breast cancer Dokmanovic M, Chang BD, Fang J, Roninson IB. (2002). Retinoid- through PPAR gamma. Mol Cell 1: 465–470. induced growth arrest of breast carcinoma cells involves co- Ou CC, Hsu SC, Hsieh YH, Tsou WL, Chuang TC, Liu JY et al. activation of multiple growth-inhibitory genes. Cancer Biol Ther (2008). Downregulation of HER2 by RIG1 involves the PI3K/Akt 1: 24–27. pathway in ovarian cancer cells. Carcinogenesis 29: 299–306. Dusek RL, Attardi LD. (2011). Desmosomes: new perpetrators in Paik J, Blaner WS, Sommer KM, Moe R, Swisshlem K. (2003). tumour suppression. Nat Rev Cancer 11: 317–323. Retinoids, retinoic acid receptors and breast cancer. Cancer Invest Esteva FJ, Yu D, Hung MC, Hortobagyi GN. (2010). Molecular 21: 304–312. predictors of response to trastuzumab and lapatinib in breast Paroni G, Brancolini C. (2011). Measurement of caspase activity: from cancer. Nat Rev Clin Oncol 7: 98–107. cell populations to individual cells. Methods Mol Biol 740: 65–79. Fitzgerald P, Teng M, Chandraratna RA, Heyman RA, Allegretto EA. Puntoni M, Decensi A. (2009). The rationale and potential of cancer (1997). Retinoic acid receptor alpha expression correlates with chemoprevention with special emphasis on breast cancer. Eur J retinoid-induced growth inhibition of human breast cancer cells Cancer 45(Suppl 1): 346–354. regardless of estrogen receptor status. Cancer Res 57: 2642–2650. Roman SD, Ormandy CJ, Manning DL, Blamey RW, Nicholson RI, Garattini E, Gianni M, Terao M. (2007). Retinoids as differentiating Sutherland RL et al. (1993). Estradiol induction of retinoic acid agents in oncology: a network of interactions with intracellular receptors in human breast cancer cells. Cancer Res 53: 5940–5945. pathways as the basis for rational therapeutic combinations. Curr Rubin I, Yarden Y. (2001). The basic biology of HER2. Ann Oncol Pharm Des 13: 1375–1400. 12(Suppl 1): S3–S8. Garattini E, Parrella E, Diomede L, Gianni M, Kalac Y, Merlini L Sheikh MS, Shao ZM, Chen JC, Hussain A, Jetten AM, Fontana JA. et al. (2004). ST1926, a novel and orally active retinoid-related (1993). Estrogen receptor-negative breast cancer cells transfected molecule inducing apoptosis in myeloid leukemia cells: modulation with the estrogen receptor exhibit increased RAR alpha gene of intracellular calcium homeostasis. Blood 103: 194–207. expression and sensitivity to growth inhibition by retinoic acid. J Gianni M, Boldetti A, Guarnaccia V, Rambaldi A, Parrella E, Raska Cell Biochem 53: 394–404. Jr I et al. (2009). Inhibition of the peptidyl-prolyl-isomerase Pin1 Shyu RY, Chang SC, Yu JC, Hsu SJ, Chou JM, Lee MS et al. (2005). enhances the responses of acute myeloid leukemia cells to retinoic Expression and regulation of retinoid-inducible gene 1 (RIG1) in acid via stabilization of RARalpha and PML-RARalpha. Cancer breast cancer. Anticancer Res 25: 2453–2460. Res 69: 1016–1026. Srinivas H, Xia D, Moore NL, Uray IP, Kim H, Ma L et al. (2006). Gianni M, Li Calzi M, Terao M, Guiso G, Caccia S, Barbui T et al. Akt phosphorylates and suppresses the transactivation of retinoic (1996a). AM580, a stable benzoic derivative of retinoic acid, has acid receptor alpha. Biochem J 395: 653–662. powerful and selective cyto-differentiating effects on acute promye- Sturniolo MT, Dashti SR, Deucher A, Rorke EA, Broome AM, locytic leukemia cells. Blood 87: 1520–1531. Chandraratna RA et al. (2003). A novel tumor suppressor protein Gianni M, Terao M, Gambacorti-Passerini C, Rambaldi A, Garattini promotes keratinocyte terminal differentiation via activation of type E. (1996b). Effects of 1,25-dihydroxy vitamin D3 on all-trans I transglutaminase. J Biol Chem 278: 48066–48073. retinoic acid sensitive and resistant acute promyelocytic leukemia Sun G, Porter W, Safe S. (1998). Estrogen-induced retinoic acid cells. Biochem Biophys Res Commun 224: 50–56. receptor alpha 1 gene expression: role of estrogen receptor-Sp1 Gianni M, Zanotta S, Terao M, Garattini S, Garattini E. (1993). complex. Mol Endocrinol 12: 882–890. Effects of synthetic retinoids and retinoic acid isomers on the Tan AR, Alexe G, Reiss M. (2009). Transforming growth factor-beta expression of alkaline phosphatase in F9 teratocarcinoma cells. signaling: emerging stem cell target in metastatic breast cancer? Biochem Biophys Res Commun 196: 252–259. Breast Cancer Res Treat 115: 453–495.

Oncogene Antitumor activity of retinoids and lapatinib G Paroni et al 3443 Terao M, Fratelli M, Kurosaki M, Zanetti A, Guarnaccia V, Paroni G signaling and inhibiting IL-6 and IL-23 receptor expression. et al. (2011). Induction of miR-21 by retinoic acid in estrogen J Immunol 181: 2277–2284. receptor-positive breast carcinoma cells: biological correlates and Yang J, Song K, Krebs TL, Jackson MW, Danielpour D. (2008a). Rb/ molecular targets. J Biol Chem 286: 4027–4042. E2F4 and Smad2/3 link survivin to TGF-beta-induced apoptosis Tsai FM, Shyu RY, Jiang SY. (2007). RIG1 suppresses Ras activation and tumor progression. Oncogene 27: 5326–5338. and induces cellular apoptosis at the Golgi apparatus. Cell Signal Yang JY, Hung MC. (2009). A new fork for clinical application: targeting 19: 989–999. forkhead transcription factors in cancer. Clin Cancer Res 15: 752–757. Tsai FM, Shyu RY, Lin SC, Wu CC, Jiang SY. (2009). Induction of Yang Q, Sakurai T, Kakudo K.. (2002). Retinoid, retinoic acid receptor apoptosis by the retinoid inducible growth regulator RIG1 depends beta and breast cancer. Breast cancer Res Treat 76: 167–173. on the NC motif in HtTA cervical cancer cells. BMC Cell Biol Yang L, Huang J, Ren X, Gorska AE, Chytil A, Aakre M et al. 10: 15. (2008b). Abrogation of TGF beta signaling in mammary carcino- Wang Y, He QY, Chen H, Chiu JF. (2007). Synergistic effects of mas recruits Gr-1+CD11b+ myeloid cells that promote metastasis. retinoic acid and tamoxifen on human breast cancer cells: proteomic Cancer Cell 13: 23–35. characterization. Exp Cell Res 313: 357–368. Zujewski J, Vaughn-Cooke A, Flanders KC, Eckhaus MA, Lubet RA, Xiao S, Jin H, Korn T, Liu SM, Oukka M, Lim B et al. (2008). Wakeﬁeld LM. (2001). Transforming growth factors-beta are not Retinoic acid increases Foxp3+ regulatory T cells and inhibits good biomarkers of chemopreventive efﬁcacy in a preclinical breast development of Th17 cells by enhancing TGF-beta-driven Smad3 cancer model system. Breast Cancer Res 3: 66–75.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)

Oncogene