Oncogene (2012) 31, 4207–4220 & 2012 Macmillan Publishers Limited All rights reserved 0950-9232/12 www.nature.com/onc ORIGINAL ARTICLE p27Kip1 represses transcription by direct interaction with p130/E2F4 at the promoters of target

R Pippa1, L Espinosa2, G Gundem3, R Garcı´a-Escudero4, A Dominguez1, S Orlando1, E Gallastegui1, C Saiz4, A Besson5, MJ Pujol1,NLo´pez-Bigas3, JM Paramio4, A Bigas2 and O Bachs1

1Department of Cell Biology, Immunology and Neurosciences, University of Barcelona, IDIBAPS, Barcelona, Spain; 2Institut Municipal d’Investigacions Me`diques-Hospital del Mar, Barcelona, Spain; 3Department of Experimental and Health Science, Research Unit on Biomedical Informatics, Universitat Pompeu Fabra, Catalan Institution for Research and Advanced Studies (ICREA), Barcelona, Spain; 4Molecular Oncology Unit, Division of Biomedicine, CIEMAT, Madrid, Spain and 5INSERM–UMR1037, Cancer Research Center of Toulouse, Toulouse, France

The cyclin-cdk (cyclin-dependent kinase) inhibitor p27Kip1 Introduction (p27) has a crucial negative role on progression. In addition to its classical role as a cyclin-cdk inhibitor, it The p27Kip1 (p27) is a cell cycle regulator whose also performs cyclin-cdk-independent functions as the main known function is the regulation of cyclin-depen- regulation of cytoskeleton rearrangements and cell dent kinase (cdk) activities (Sherr and Roberts, 1999). The motility. p27 deficiency has been associated with tumor importance of p27 as a cell cycle regulator in vivo was aggressiveness and poor clinical outcome, although the revealed by the generation of p27À/À mice that display an mechanisms underlying this participation still remain increase in body size and multiple organ hyperplasia (Fero elusive. We report here a new cellular function of p27 as et al., 1996; Nakayama et al., 1996) (Kiyokawa et al., a transcriptional regulator in association with p130/E2F4 1996). This protein was originally considered as a complexes that could be relevant for tumorigenesis. We ‘universal’ inhibitor of cyclin-cdk complexes whose observed that p27 associates with specific promoters of activity was exerted by the N-terminal half of the molecule genes involved in important cellular functions as proces- (Russo et al., 1996). However, recent reports revealed a sing and splicing of RNA, mitochondrial organization and dual role of p27, behaving as an inhibitor or and activator respiration, and cell cycle. On these promoters of cyclin D-cdk4/6 and cyclin E-cdk2 on depending p27 co-localizes with p130, E2F4 and co-repressors as of specific tyrosine phosphorylation (Chu et al., 2007; histone deacetylases (HDACs) and mSIN3A. p27 co- Grimmler et al., 2007; Blain, 2008; James et al., 2008). immunoprecipitates with these and by affinity p27 deficiency is associated with tumorigenesis. Re- chromatography, we demonstrated a direct interaction of duced p27 levels are frequently observed in human cancers p27 with p130 and E2F4 through its carboxyl-half. We in association with tumor aggressiveness and poor clinical have also shown that p130 recruits p27 on the promoters, outcome (Slingerland and Pagano, 2000). In most of the and there p27 is needed for the subsequent recruitment of cases, low levels of p27 are the consequence of an HDACs and mSIN3A. Expression microarrays and increased protein degradation (Frescas and Pagano, luciferase assays revealed that p27 behaves as transcrip- 2008). In some tumors, cytoplasmic localization of p27 tional repressor of these p27-target genes (p27-TGs). is also associated with increased motility, aggressiveness Finally, in human tumors, we established a correlation and poor prognosis (Viglietto et al., 2002). with overexpression of p27-TGs and poor survival. Thus, Quiescent cells contain high levels of p27 mostly this new function of p27 as a transcriptional repressor located in the nucleus. After mitogenic stimulation, it could have a role in the major aggressiveness of tumors translocates to the cytoplasm where it is degraded via with low levels of p27. ubiquitin–proteasome (Shirane et al., 1999). In the Oncogene (2012) 31, 4207–4220; doi:10.1038/onc.2011.582; cytoplasm, p27 performs several cdk-independent activ- published online 19 December 2011 ities. It participates in both actin cytoskeleton rearran- gement and cell motility through the modulation of Keywords: p27; p130; E2F4; transcription RhoA activity, a role that relies in the C-terminal half of p27 (McAllister et al., 2003; Besson et al., 2004). p27 also interacts with stathmin, a microtubule-associated protein, thus regulating cell morphology and motility Correspondence: Professor O Bachs, Department of Cell Biology, (Baldassarre et al., 2005). These regulatory roles of p27 Immunology and Neurosciences, Faculty of Medicine, University of on cell migration have been associated with the cdk- Barcelona, Institut d’Investigacions Biome` diques August Pi i Sunyer independent oncogenic functions of cytoplasmic p27 (IDIBAPS), Casanova 143, 08036-Barcelona, Spain. E-mail: [email protected] (Besson et al., 2008). Received 15 August 2011; revised 29 October 2011; accepted 11 It is assumed that the main function of nuclear p27 November 2011; published online 19 December 2011 during early G1 is to prevent premature entry into S Transcriptional regulatory role of p27Kip1 R Pippa et al 4208 phase by maintaining cyclin E/cdk2 complexes inactive TRANSFAC database, we found that target sequences (Chu et al., 2008). However, whether nuclear p27 in of ETS1, E2F4 and GABP were significantly enriched in quiescent cells solely acts as a CDK inhibitor or may the promoters of these genes (Figure 2a). We subse- fulfill other nuclear functions remains unclear. It has quently analyzed whether p27-TGs were also targeted been postulated that p27 may participate in the by these three mentioned TFs (Hollenhorst et al., 2007; regulation of transcription independently of cyclin-cdk Lopez-Bigas et al., 2008). Results revealed that p27 regulation. It has been shown that the interaction of p27 shared a significant number of target genes with these with neurogenin-2 leads to the differentiation of neural TFs and additionally also with ELF1 and RBP2 progenitors in the cortex. Specifically, p27 stabilizes (Figure 2b and web Supplementary analysis S1). neurogenin-2 by a mechanism that depends on the To analyze the putative interaction of p27 with E2F4, integrity of its N-terminal half but does not require we performed immunoprecipitation (IP) experiments interactions with cyclins and cdks (Nguyen et al., 2006). with anti-p27 and observed that E2F4 co-immunopre- Moreover, the overexpression of p27 in C2C12 cells cipitated with p27 (Figure 2c). This interaction was induces myogenic differentiation, whereas elimination of confirmed by IPs with anti-E2F4 in mice embryo p27 prevents differentiation (Munoz-Alonso et al., fibroblasts (MEFs) p27WT or p27À/À (Supplementary 2005). In another example it has been shown that Figures S2a and b). Moreover, by ChIP, we demon- overexpression of p27 induces the expression of ery- strated that p27 and E2F4 simultaneously associated throid markers in the K562 cell line (Acosta et al., 2008). with several p27-TG promoters enriched with E2Fs- All these data suggested that the nuclear role of p27 in binding sequences (Supplementary Figures S3 and S4) quiescence might rely not only on the inhibition of but not with the control (HoxB8) (Figure 2d). We cyclin-cdk complexes but also on the transcriptional subsequently studied whether p27 interacted with repression of specific target genes. specific co-repressors known to be associated with Thus, we hypothesize that p27 could interact with E2F4. IP experiments revealed that p27 co-immunopre- transcriptional regulators at specific promoters. We cipitated with endogenous p130, mSIN3A and histone explored this possibility by chromatin immunoprecipita- deacetylases (HDACs) 1, 4 and 5, indicating that they tion (ChIP) analysis followed by promoter microarrays form complexes in vivo (Figure 2e). The interaction of (chip). Results revealed that p27 associates with a p27 with p130 was confirmed by IPs with anti-p130 number of promoters in quiescent cells. These p27 (Supplementary Figure S2c). Finally, we checked target genes (p27-TGs) are mainly involved in RNA whether all these repressors associated with p27 on processing, translation, cell cycle and respiration. We p27-TG promoters. As observed in Figure 2f, in found that p27 has a role as a transcriptional repressor quiescent cells p27 associated with p130, mSIN3A and in coordination with p130 and E2F4. We have shown HDACs 1, 4 and 5 on several p27-TG promoters but not that p27-TGs are overexpressed in different tumors and on the control gene promoter. We also observed that this is associated with poor clinical outcome. p27 co-immunoprecipitated with ETS1 (Figure 2g). Moreover, ChIP experiments revealed the simultaneous presence of p27 and ETS1 on some p27-TG promoters Results enriched with Ets-binding sites (Figure 2h and Supple- mentary Figure S5). p27 binds to specific gene promoters Quiescent cells contain high levels of p27 mostly in the p130 directly interacts with p27 and recruits it nucleus (Supplementary Figure S1) (Shirane et al., on the p27-TG promoters 1999). Whether p27 has a role, aside from cyclin-cdk To analyze the functional relationship between p27 and inhibition in the nucleus of quiescent cells still remains p130/E2F4 in transcriptional regulation, we performed unclear. We therefore tested the possibility that p27 IP experiments using anti-p27 on extracts from p27wt, could associate with specific gene promoters to regulate p27À/À and p130À/À MEFs and additionally on extracts transcription. ChIP on chip experiments performed in from two MEF-lines harboring a p27 mutant containing quiescent cells revealed that p27 significantly associates a deletion of the first 51 aa (p27D51) or four punctual with a number of gene promoters (Figure 1a and web mutations that unable its interaction with cyclins and Supplementary dataset S1). The interaction of p27 with cdks (p27CKÀ; Supplementary Figure S6). Results some of these promoters was validated by real time revealed that p27wt, p27D51 and p27CKÀ interacted with qPCR of immunoprecipitated chromatin fragments p130 (Figure 3a), indicating that this association is (Figure 1b). enrichment analysis indi- cyclin-cdk independent. Then, we studied whether p130 cates that p27 target genes (p27-TGs) are mostly was able to directly interact with p27. Thus, affinity involved in pre-mRNA processing and splicing, mito- chromatography experiments using two fragments of chondrial biogenesis and respiration, translation and p27: the N-terminal half (aa 1–110) (p27-NT) and the cell cycle (Figure 1c). C-terminal half (aa 110-198) (p27-CT), coupled to sepharose 4B were performed. Two different fragments p27 associates with specific transcription factors (TFs) of p130: p130NT (aa 1-417) and p130CT (aa 967-1082) on the p27-TG promoters were separately loaded onto both p27 columns. Results We then analyzed the promoter sequences of the p27- showed that fragment p130CT, but not p130NT, TGs for the presence of TF-binding sites. Using the directly interacted with p27-CT on the column

Oncogene Transcriptional regulatory role of p27Kip1 R Pippa et al 4209

a p27 IgG 4

2 M

0

-2

6 810121416 A b 60

40

20 Relative binding

0 Ing4 Rbl2 Kif11 Eif2a Sfrs8 Coq5 Med6 Clcn7 Kntc1 Aurka Ccnt2 Copb1 control Heatr1 Med18 Fancd2 Anpc13 Cdc123 Gemin5 Gorasp2

# genes in c Go category Observed Expected Pvalue category Biological process mRNA processing 55 17 1.85E-14 234 RNA splicing 51 14 8.90E-16 190 translation 77 35 1.80E-10 493 Mitochondrion organization and 13 4 1.60E-04 61 biogenesis cytokinesis 10 3 1.40E-04 39 biogenesis and assembly 16 6 9.60E-05 61 cell division 29 16 6.80E-04 220 Golgi vesicle transport 14 4 5.70E-06 51 Molecular function RNA binding 85 39 6.20E-11 555 NADH dehydrogenase activity 20 4 1.40E-10 49 translation regulator activity 29 8 5.80E-09 119 structural constituent of ribosome 26 13 5.40E-04 187 Cellular component ribosome 29 14 1.30E-04 198 spliceosome 28 9 6.80E-08 126 mitochondrion 98 57 2.90E-07 810 mitochondrial membrane 45 18 3.50E-08 259

Figure 1 p27 associates with specific gene promoters in quiescent NIH 3T3 cells. (a) DNA was immunoprecipitated with anti-p27 or IgG and then, labeled with Cy5 and Cy3, respectively, whereas input DNA remained unlabeled. Both samples were hybridized in two different arrays, one with p27 vs input and the other one with IgG vs input. This plot represents the log ratio of channels (Cy3 and Cy5) as M value (M ¼ log2(Cy3)-log2(Cy5)) vs the intensity of each spot in the array (A ¼ (log2(R) þ log2(G))/2) as log of geometric average of the channels intensities for each spot (Cy3 ¼ red, Cy5 ¼ green). (b) ChIP was performed with anti-p27 and then real time qPCR was performed using specific primers for the 19 mentioned gene promoters. Results were normalized vs the values obtained with the IgG control and represented as the relative binding. (c) p27-TGs were analyzed for enrichment of GO terms using binomial distribution in Gitools. This test measures whether the proportion of genes with particular GO term is significantly higher than random expectations. P-values are shown in a color code scale; red color shows significant P-values after multiple testing correction, indicating enrichment for a particular GO term. Gray means no statistically significant difference.

Oncogene Transcriptional regulatory role of p27Kip1 R Pippa et al 4210 observed expected N p-value ETS 192 150 10075 4.90E-06 E2F4 44 30 1777 5.30E-06 PREDICTION GABP 61 27 839 3.50E-14

observed expected N p-value ETS 62 20 817 <1E-16 E2F4 41 18 738 7.8E-11 GABP 57 22 894 <1E-16 CHIP-ON-CHIP ELF1 73 28 1127 <1E-16 RBP2 50 11 437 <1E-16

ChIP Input p27 E2F4 IgG IP Ccnt2 Input IgG p27 E2F4 Gorasp2

Kntc1 p27 Sfrs8 IP HoxB8 Input IgG p27

p130 ChIP

p27 input p27 c19 p27 ac p130 Hdac1 Hdac4 Hdac5 mSin3A IgG

mSIN3A Gorasp2

p27 Kntc1

HDAC1 Sfrs8

p27 HoxB8

IP HDAC4 Input IgG p27 p27 ETS1

HDAC5 p27

p27 ChIP Input p27 ETS1 IgG HDAC2 Clcn7

p27 Copb1

HoxB8

Figure 2 p27 associates with E2F4-p130 complexes and with ETS1 on the promoters of p27-TGs. (a) Identification of the enrichment of predicted transcription factors-binding sites (TFBS) in p27-TG promoters. The P-values are shown in a color code scale; red shows significant P-values, whereas gray means no statistically significant difference. (b) Enrichment of experimentally validated target genes for several transcription factors among p27-TGs. In all cases, these analyses were performed by Poisson statistics as described under material and methods. (c) IP experiments were performed in NIH 3T3 quiescent cells using anti-p27 or non-specific IgGs (used as a control). The presence of p27 and E2F4 in the immunoprecipitates was detected by WB. (d) ChIP experiments using antibodies against E2F4 and p27 and a non-specific IgG were performed on four p27-TG promoters and on a control promoter (HoxB8). consensus sites in these promoter sequences are shown in the Supplementary Figures S3 and S4. (e) IP experiments were performed using anti-p27 or non-specific IgGs (used as a control). Then, the immunoprecipitates were analyzed for the presence of p130, mSIN3A, HDAC1, 4 and 5 by WB. (f) ChIP experiments were performed to determine the presence of repressors mentioned in (e) on the promoters of the mentioned p27-TGs and on a control promoter (HoxB8). (g) IP experiments were performed using anti-p27 or non-specific IgGs (used as a control). The presence of p27 and ETS1 in the immunoprecipitates was detected by WB. (h) ChIP experiments using antibodies against ETS1, p27 and a non-specific IgG were performed on two p27-TG promoters and on a control promoter (HoxB8). ETS consensus sites in these promoter sequences are shown in the Supplementary Figure S5.

Oncogene Transcriptional regulatory role of p27Kip1 R Pippa et al 4211 (Figure 3b). Similar experiments were performed by aimed to study the p27-E2F4 interaction in p130À/À loading purified E2F4 on both p27 columns. Results MEFs. IP experiments indicate that p130 is needed to revealed that p27-CT but not p27-NT was able to form the p27-E2F4 complexes in the cells (Figure 3f). directly associate with E2F4 (Figure 3c). Next, we studied whether p27 needs p130 to be p27 is necessary to recruit different co-repressors on p27- recruited on the promoters. Thus, ChIP experiments TG promoters wt CKÀ D51 using anti-p27 were performed on p27 , p27 , p27 , We next explored whether p27 is needed on the À/À À/À p27 and p130 MEFs. Results revealed that all these promoters to recruit co-repressors as mSIN3A or wt CKÀ D51 forms of p27 (p27 ,p27 and p27 ) significantly HDAC1. Thus, ChIP was performed with anti-p27, associated with p27-TGs promoters when p130 was anti-HDAC1 and anti-mSIN3A on p27wt and p27À/À present in the cells supporting the cyclin-cdk-indepen- MEFs. Then, the association of these proteins with the dent interaction of p27 with promoters. In contrast, p27 Heatr1 and RBl2 promoters was analyzed by qPCR. À/À did not associate with the promoters in p130 MEFs, Results showed that in the p27À/À MEFs the association indicating that p27 needs p130 to be recruited on the of HDAC1 and mSIN3A with the Heatr1 promoters promoters (Figure 3d). Interestingly, ChIP experiments (Figure 4a and Supplementary Figure S7) and with the using anti-p130 showed that this protein interacted with RBl2 promoter (Figure 4b and Supplementary Figure promoters independently of p27 (Figure 3e). Finally, we S7) was strongly reduced, indicating that p27 is necessary to recruit these co-repressors. Interestingly,

Inputs IP p27 IP IgG

p27 p27 p27 p130 p27 p27 p27 p27 p130 p27 p27 p27 p27p130 p27 Cells wt ck Δ51 kO kO wt ck Δ51 kO kO wt ck Δ51 kO kO Heatr1 Rbl2 p130 p27 HDAC1 mSIN3A 20 p27 HDAC1 mSIN3A 8 Cdk4 6 15 CycD1 4 10 p27 2 5 p27Δ51 0 0 Chromatin enrichment Chromatin enrichment p27 WT p27 KO p27 WT p27 KO GST GST-p27 GST GST-p27 wt -/- wt -/- Input UB W B UB W B p27 p27 p27 p27 p27NT Input UB W B UB W B p130 Heatr1 Rbl2 p27 NT NT p27CT E2F4 Actin Actin p130 p27NT p27 CT CT p27CT Heatr1 Rbl2 10 p130 E2F4 ETS1 30 p130 E2F4 ETS1 4 4 20 3 3 5 2 2 10

1 1 0 0 Chromatin enrichment p27 WT p27 KO Chromatin enrichment p27 WT p27 KO Chromatin enrichment Chromatin enrichment 0 0 p27 p27 p27 p130 p27 p27 p27 p27 p130 p27 IP WT CK d51 KO KO WT CK d51 KO KO Input IgG p27 IP p130 Input IgG p27 HDAC1 p130

E2F4 mSIN3A

p27 p27

Figure 3 p130 targets p27 to the promoters. (a) Immunoprecipita- Figure 4 p27 recruits co-repressors as HDAC1 and mSIN3A on tion of p27 in p27wt, p27CKÀ, p27D51, p130À/À and p27À/À MEFs were the promoters. (a, b) Graphical representation of chromatin performed to evaluate the interaction of cyclin D1, CDK4 and enrichment obtained in ChIP experiments using p27, HDAC1 or p130 with the different variants of p27. The presence of these mSIN3A antibodies in quiescent p27WT and p27À/À MEFs on proteins in the immunoprecipitates was determined by WB. Heatr1 (a) and Rbl2 (b) promoters. Data are represented as the (b, c) Affinity chromatography columns containing p27-NT (aa mean value±s.d. of three independent experiments. (c, d) Protein 1-110) or p27-CT (aa 110-198) were loaded with purified p130NT levels of Heatr1 (c) and Rbl2 (d) in p27WT and p27À/À MEFs as and p130CT (b) or E2F4 (c). After extensive washing the unbound determined by WB. Levels of actin were used as a loading control. protein (UB), the last wash (W) and the bound protein was (e, f) Graphical representation of chromatin enrichment obtained measured by WB. (d, e) Graphical representation of chromatin in ChIP experiments using p130, E2F4 or ETS1 antibodies in enrichment obtained in ChIP experiments using p27 (d) or p130 (e) quiescent p27WT and p27À/À MEFs on Heatr1 (e) and Rbl2 (f) antibodies and performed in p27wt, p27CKÀ, p27D51, p130À/À and promoters. Data are represented as the mean value±s.d. of three p27À/À MEFs. (f) IP experiments on p130À/À MEFs were performed independent experiments. (g) IP experiments on p130À/À MEFs using anti-p27 or non-specific IgGs (used as a control). The were performed using anti-p27 or non-specific IgGs (used as a presence of p27, p130 and E2F4 in the immunoprecipitates was control). The presence of p130, p27, HDAC1 and mSIN3A in the detected by WB. immunoprecipitates was detected by WB.

Oncogene Transcriptional regulatory role of p27Kip1 R Pippa et al 4212 the levels of the proteins encoded by these two genes are p27wt cells, indicating that p27 represses the expression clearly increased in p27À/À cells (Figures 4c and d), of these genes (Figure 5a and web Supplementary suggesting that p27 could act as a transcriptional co- datasets S2 and S3). repressor. Figures 4e and f show that in p27À/À cells p130 We also studied the cdk dependence of the p27- and E2F4 still remain associated with these promoters. mediated transcriptional repression, by expression Finally, we observed that in the absence of p130, p27 microarrays using p27CKÀ MEFs (Besson et al., 2007). still retains the ability to interact with both HDAC1 and Results showed that similar to that observed in p27À/À mSIN3A co-repressors (Figure 4g). cells, a significant number of p27-TGs genes was upregulated in p27CKÀ cells (Figures 5b and c), indicat- p27 is a transcriptional co-repressor ing that a group of genes is repressed by p27 in a cyclin- To analyze the role of p27 as a transcriptional regulator, cdk-dependent manner. Consistent with the idea that we performed expression microarray analysis on quies- cyclin-cdk complex inhibition is needed to repress the cent p27wt and p27À/À MEFs. Results showed that a expression of this specific subset of p27-TGs, a significant number of genes were differentially expressed significant number of these genes were overexpressed in p27À/À vs p27wt cells. Among them, a significant group in both, p27CKÀ and p27À/À MEFs (Figure 5d, upper of p27-TGs showed an increased expression in p27À/À vs panel). Interestingly, we also observed that a significant

Decrease Increase Decrease Increase

CK- vs WT KO vs WT 3239 3460 4214 4283

114 118 52 99 195 276 exp p-value exp p-value exp p-value exp p-value 88 0.00189 67 0.98 targets 69 4.22E-05 targets 86 1.78E-04

CK up KO up 118 99 Decrease Increase

exp p-value CK- vs KO 4324 4189 32 0.002 69 50 49

132 101 CK down KO down 114 52 194 exp p-value exp p-value exp p-value targets 87 2.49E-07 85 0.029 39 0.95 84 22 30

Decrease Increase

Δ51 vs WT DOWN-REGULATION UP-REGULATION 3039 4463

73 130 51 vs. WT 51 vs. WT Δ CK vs. WT CK vs. KO KO vs. WT Δ CK vs. WT CK vs. KO KO vs. WT 218 exp p-value exp p-value ELF1 65 0.16 targets 96 9.35E-05 GABP ETS Δ51 up CK up RUNX 130 118 RBP2

exp p-value p130 in G0 85 73 26 3.59E-04 E2F4 in G0 45 Figure 5 Deregulation of in p27-deficient and p27-mutated MEFs. (a) Comparative expression microarray analysis was performed in p27wt- and p27À/À MEFs. The Venn diagram shows the genes that are upregulated (red) or downregulated (dark blue) in the p27À/À MEFs (KO) vs p27wt cells. The expression of some p27-TGs was unmodified (yellow), whereas some others were overexpressed (orange) or downregulated (pale blue). P-values are shown in a color code scale; red color shows significant P-values, whereas gray means no statistically significant difference. (b, c) Venn diagrams representing the comparative expression analysis in p27CK vs p27wt cells (b) and p27CKÀ- vs p27À/À MEFs (c). (d) Diagrams showing the overlap of upregulated (upper panel) or downregulated (bottom panel) p27-TGs in p27CKÀ and p27À/À MEFs. (e) Venn diagrams of comparative expression analysis obtained in p27D51 vs p27wt MEFs. (f) Diagrams showing the overlap of upregulated p27-TGs in p27D51 and p27CKÀ MEFs. (g) Differentially expressed genes in p27CKÀ, p27D51 and p27À/À were analyzed for enrichment of GO terms using binomial statistics in Gitools. This test measures whether the proportion of genes with a particular GO term differs statistically from random expectations. P-values are shown in a color code scale; red color shows significant P-values after multiple testing correction, indicating enrichment for a particular GO term. Gray means no statistically significant difference.

Oncogene Transcriptional regulatory role of p27Kip1 R Pippa et al 4213 number of p27-TGs were downregulated in p27CKÀ clearly blocked the expression of both genes (Figures 6c MEFs, indicating that p27 could also repress transcrip- and d). tion in a cdk-independent manner (Figure 5b and web To finally corroborate the transcriptional repressor Supplementary datasets S4 and S5). The ability of role of p27, we generated different constructs p27CKÀ to repress transcription is highlighted when containing a luciferase reporter gene under the control expression data from p27CKÀ versus p27À/À MEFs were of p27-TG promoters including Aurka, Coq5 and compared (Figure 5c and web Supplementary datasets Sfrs8. The promoter of Noxa (a non-p27-TG) was used S6 and S7). Although a group of p27-TGs were as a control. Luciferase assays revealed that knockdown downregulated in both p27CKÀ and p27À/À MEFs, this of p27 induced a transcriptional activation of the was not statistically significant (Figure 5d, bottom p27-TGs but not that of the control gene (Figures 6e panel). and f). Altogether these results reveal a transcriptional repressor activity of p27 that is performed either in a Overexpression of p27-TGs in tumors correlates with a cdk-independent or in a cdk-dependent manner depend- poor survival ing on the subset of genes regulated. It is known that tumors with reduced p27 levels display To analyze the domains of p27 involved in transcrip- a worse outcome (Slingerland and Pagano, 2000). Thus, tional regulation, expression microarray analysis on we can speculate that the higher malignancy of these p27D51 versus p27WT MEFs was performed. The protein tumors could be due to increased expression of some of p27D51 lacks the 51 N-terminal amino acid, a region the hereby-identified p27-TGs. To test this possibility, that contains the interacting domains of p27 with cyclins we analyzed the expression status of p27-TGs in a and cdks (Kiyokawa et al., 1996). Similarly to p27À/À collation of oncogenomic experiments from IntOGen, a and p27CKÀ MEFs, a significant number of p27-TGs publicly available cancer resource that contains infor- were upregulated in these cells (Figure 5e and web mation on the pattern of different alterations in diverse Supplementary datasets S8 and S9), as being a types of cancer (Gundem et al., 2010). As shown significant number of them also overexpressed in in Table 1 and web Supplementary analysis S2, a p27CKÀ MEFs (Figure 5f). These data support that significant number of p27-TGs were overexpressed in inhibition of cdk activity is needed for repression of different types of human tumors. We next analyzed the these genes. However, in contrast to results observed in relationship between reduced p27 expression with p27CKÀ cells no significant repression of p27-TGs was clinical parameters, using a breast cancer tissue micro- observed in p27D51 cells (Figure 5e). These results array containing tumors of different origins and grades. indicate that p27D51 is not as efficient as p27CKÀ as a Reduction of p27 levels correlated with an increased repressor, and suggest that a domain in the first 51 aa of grade of the tumors (Figure 7A). Next, we tested the p27 is also needed for the cdk-independent gene correlation between the levels of Aurka and Fancd2 repression. (two of the most commonly represented p27-TGs We also performed enrichment analysis with targets among different samples in the IntOGen database) and of p130, E2F4 and members of the Ets family of TFs. the expression of p27. We observed that reduced levels Interestingly, the p27-TGs upregulated in p27À/À cells of p27 were significantly linked to increased nuclear were also targets of the Ets-TFs but not of p130 and expression of Fancd2 and Aurka (Figures 7B and C). Of E2F4 (Figure 5g lane 8). These results suggest that p27 note, normal breast samples included in the array collaborates with Ets family in the transcriptional showed clear nuclear presence of p27 (Figure 7B-a) repression of this particular set of p27-TGs. concomitant with almost undetectable Fancd2 and Moreover, the p27-TGs upregulated in p27CKÀ and Aurka (Figure 7B-b and 7B-c). A multivariate analysis p27D51 cells are also targets of Ets-TFs but they are revealed that low levels of p27 correlate with tumor additionally enriched among targets of p130 and E2F4 grade and with increased levels of Aurka and Fancd2 (Figure 5g, lanes 5 and 6). Finally, results also revealed but not with the expression of PCNA, a proliferation that the p27-TGs repressed in p27CKÀ MEFs were targets marker (Supplementary Table S1). Finally, we searched of Ets-TFs (Figure 5g lane 2). Collectively, these data for possible overlapping between p27-TGs overex- suggest that p27 participates in transcriptional regula- pressed in human tumors with patient prognosis in the tion through at least two different complexes one Oncomine gene expression database (Rhodes et al., containing Ets-TFs and another one containing p130/ 2004). One study of breast cancer (Pawitan et al., 2005) E2F4. and one study of astrocytoma (Phillips et al., 2006) with To further confirm the role of p27 as a transcriptional very significant association (P-valueo10À4) were se- co-repressor, we analyzed the effect of re-introduction of lected for further Kaplan–Meier analysis. In addition, a p27 in p27À/À MEFs on the expression of two p27-TGs large study of lung adenocarcinomas, including samples upregulated in these cells. We observed that tranfection from four different clinical research centers, was of p27 repressed the expression of both genes (Figures analyzed (Shedden et al., 2008). In all these cases, a 6a and b and Supplementary Figure S7). Similar significant association between p27-TGs overexpression experiments were carried out in NIH 3T3 cells, in which and poor survival was observed (Figure 7D), suggesting endogenous p27 was knocked down with a specific that the functional loss of p27 repression of p27-TGs shRNA. In these cells both genes were upregulated, but could be associated with higher metastatic capacity of the overexpression of a non-degradable form of p27 primary tumors.

Oncogene Transcriptional regulatory role of p27Kip1 R Pippa et al 4214

control p27shRNA + ø vector 5 p27shRNA p27shRNA + p27WT 4 p27 WT p27 KO p27 KO +empty p27 KO + 3 p27 2

Expression fold 1 Actin 0 Heatr1 Rbl2

control p27shRNA + ø vector p27shRNA p27shRNA + p27WT 4 Control ShRNA p27l ShRNA p27l ShRNA + CFP p27l ShRNA + CFPp27

3 CFP p27

2 p27

1 Expression fold Actin 0 Heatr1 Rbl2

3 control p27shRNA

p27 2 Actin 1 control shRNA + - p27 shRNA - +

Rel. luciferase activity 0 Pgl2b Aurka Coq5 Sfrs8 Noxa Figure 6 p27 represses transcription of p27-TGs. (a) The levels of mRNA for two specific p27-TGs (Heatr1 and Rbl2) were measured by qPCR in quiescent MEFs with p27WT (black bar), p27KO (white bar) and p27KO infected with empty vector (vertical lines bar) or p27WT (gray bar). Data are represented as the mean value±s.d. of three independent experiments. (b) WB of p27 levels in cells used in (a). (c) Expression levels of Heatr1 and Rbl2 were measured by qPCR in quiescent NIH 3T3 transfected with control shRNA (black bar), p27 shRNA (white bar) and NIH 3T3 with p27 shRNA plus CFP (vertical lines bar) and p27 shRNA plus CFP-p27 (gray bar). Data are represented as the mean value±s.d. of three independent experiments. (d) WB of p27 levels in cells used in (c). (e) Quiescent control and p27-knocked down NIH 3T3 cells were transfected with specific reporter constructs. Results show the luciferase activity in p27-knocked down cells (grey bars) relative to the activity in control cells (black bars) for p27-TGs promoters and for the control promoter Noxa. Data are represented as the mean value±s.d. of three independent experiments. (f) p27 levels in cells of experiments in (e) as determined by WB.

Table 1 p27-TGs are over-expressed in many types of cancer Reference Authors’ year Topography Observed Expected NP-value

GSE6791 Pyeon et al., 2007 Cervix uteri 64 43 404 7.79E-04 GSE6338 Piccaluga et al., 2007 Blood 95 70 404 7.85E-04 GSE8671 Sabates-Bellver et al., 2007 Colon 43 30 404 1.26E-02 GSE6691 Gutie´rrez et al., 2007 Blood 55 43 317 2.95E-02 GSE7904 Richardson et al., 2006 Breast 62 49 404 3.56E-02 GSE10797 Casey et al., 2009 Breast 30 21 317 4.17E-02 GSE1037 Jones et al., 2004 Lung 64 51 360 4.33E-02 GSE12453 Brune et al., 2008 Lymph nodes 72 59 404 4.47E-02

Statistics of p27-TGs in different IntOGen data sets. Each row gives the result of the test comparing the genes upregulated in a specific data set studying the indicated number of cancer samples and the p27-TGs. See Methods for details.

Discussion through cdk-independent functions (Sicinski et al., 2007). However, the specific mechanisms underlying Compelling evidence unveiled a dual role for p27 during this participation still remain elusive. We report here a tumorigenesis: as a tumor suppressor presumably by new function of p27 as a transcriptional regulator that virtue of its cdk regulatory function and as an oncogene can have a role during oncogenesis. We observed that

Oncogene Transcriptional regulatory role of p27Kip1 R Pippa et al 4215 ABabp27 FancD2c Aurk A 1 (n=9) 0.9 p27 cdkn1b Normal Breast 0.8 d e f 0.7 (n=24) 0.6 (n=22) 0.5 Tumor grade II

0.4 g h i (n=6) 0.3 (n=33)

0.2 Tumor grade III

Relative p27 expression (arbitrary units) 0.1 j kl 0 sis Lymph Node etasta al breast) or gradeor Igradeor gradeII III metastasis Tum Tum Tum ph Node m Control (Norm Lym

C 2 2 R=-0.913 R=-0.918 p<0.0001 p<0.0001 AurkA Staining Score (A.U) 0 FanCD2 Staining Score (A.U) 0

0.0 0.5 1.0 1.5 0.0 0.5 1.0 1.5 p27 Staining Score (A.U.) p27 Staining Score (A.U.)

D 1.0 1.0 1.0

0.8 0.8 0.8

0.6 0.6 0.6 urvival urvival urvival

0.4 0.4 0.4 Overall s Overall s Overall s

0.2 0.2 0.2

0.0 0.0 0.0 0 2 4 6 8 0 1231 2 3 0 1 2 3 years years years n=179 n=77 n=442 p-val=1.5x10-5 p-val=4.6x10-4 p-val=6.1x10-6 Pawitan Breast Cancer Phillips Astrocytoma Beer Lung Adenocarcinoma Figure 7 Overexpression of p27-TGs correlates with poor survival in human cancers. (A) Score of p27 staining in correlation with the histopathological grade of a tissue microarray of breast cancer. (B) Representative staining of p27 (a, d, g, j), Fancd2 (b, e, h, k) and Aurka (c, f, i, l). Samples were: normal breast tissue (a, b, c), ductal breast carcinoma Grade II (d, e, f), grade III (g, h, i) and lymph node metastasis (j, k, l). Bar ¼ 150 mm. (C) Graphic representation of the correlation between the expression of p27 and that of Aurka and Fancd2 in tissue microarrays. (D) Kaplan–Meier analyses were performed on a number on tumors from breast, astrocytoma and lung adenocarcinoma. p27 associates with promoters of genes involved in tional repressors of genes necessary for DNA replication cellular functions as processing and splicing of RNA, during G0/G1 (Macaluso et al., 2006). At late G1 cyclin mitochondrial organization and respiration, translation D-Cdk4/6 and subsequently cyclin E-Cdk2, phosphor- and cell cycle. The expression of a group of these p27- ylate p130 disrupting these complexes and allowing TGs is regulated by members of the Ets family of TFs, transcriptional activators, , to induce the expres- whereas another group of p27-TGs is regulated by p130/ sion of these genes (Malumbres and Barbacid, 2005). E2F4 complexes. We investigated the functional relationship between p27 The p130/E2F4 complexes are crucial in the regula- and p130/E2F4 complexes, and we have shown that p27 tion of G1/S transition. They are the major transcrip- associates with most of the proteins of these complexes

Oncogene Transcriptional regulatory role of p27Kip1 R Pippa et al 4216

1 2 3

mSin3A N N HDAC

p130 p27 p130 p27 p130

E2F4 E2F4 E2F4 C C

p27-TG promoter p27-TG promoter p27-TG promoter

No Transcription No Transcription No Transcription Figure 8 Model illustrating the participation of p27 on the organization of p130/E2F4 repressor complexes. p130 first drives E2F4 to the promoters (1), then p27 is subsequently loaded by directly interacting by its carboxyl-domain with both p130 and E2F4 (2), finally, p27 recruits the co-repressors HDAC1 and mSIN3A on these promoters (3).

and that they co-localize on p27-TG promoters in tional activation of the p130/E2F4 regulated genes the quiescent cells. The observation that p27 directly presence of p27 on the promoters, but also of active interacts with p130 and E2F4 through its carboxyl-half cyclin-cdk complexes, are concomitantly needed. All corroborates that p27 forms part of these repressor these data lead us to postulate a model in which in complexes. It is known that p130 is essential to transport quiescent cells p27 participates in the organization of the E2F4 to these promoters (Rayman et al., 2002). Our p130/E2F4 complexes to repress the genes needed for results indicate that p130 is necessary also for the DNA replication. At mid-late G1 when the expression of recruitment of p27 on the promoters and that this is a these genes is needed, the p27 associated with p130/ cyclin-cdk-independent process. Interestingly, we also E2F4 on the promoters would recruit cyclin-cdk observed that p27 is critical for the subsequent assembly complexes by its N-terminal domain. This model is of HDAC1 and mSIN3A in these complexes. Thus, a compatible with our data showing that p27 directly model appears in which p130 drives p27 and E2F4 to the binds to p130 and E2F4 through its carboxyl-half; thus, promoters and there, p27 is responsible for the recruit- under these conditions, the NH2-half of p27 is ‘free’ to ment of the other co-repressors (Figure 8). interact when necessary with cyclin and cdk molecules. Our results also revealed that p27 behaves as a The model also postulates that these cyclin-cdk com- transcriptional repressor. This was demonstrated by plexes bound to p27 on the promoters would be expression microarray analysis performed on p27À/À subsequently activated and then they could phosphor- MEFs that allowed us the identification of p27-TGs ylate p130 and thus disrupting the repressor complexes. that were upregulated in these cells. This upregulated So, our model postulates that the binding of p27 to expression could be repressed by the re-introduction of promoters, in addition to participate in the transcrip- p27 in these knockout MEFs but also in NIH 3T3 cells tional repression, could help to recruit cyclin-cdk knocked down for p27. The transcriptional repressor complexes needed for p130 phosphorylation. Work role of p27 was also supported by assays carried out on trying to validate this model is currently performed in cells containing a luciferase reporter gene under the our laboratory. control of p27-TG promoters. Finally, expression A corollary of these data is the existence of over- microarray analysis performed on p27CKÀ MEFs re- lapping functions for p130 and p27. Indeed, the vealed that p27CKÀ is a good transcriptional repressor inactivation of p107 and p27D51 in vivo (Yeh et al., and indicated that p27 mediates the transcriptional 2007) is highly reminiscent of that of p130 and p107 repression in a cdk-independent manner. (Cobrinik et al., 1996). However, in vivo unique Interestingly, as shown in Figure 5e lane 2, the p27- functions for p27 and p130 also exist (Soeiro et al., TGs repressed in p27CKÀ MEFs were only those 2006). Whether these unique functions can be related to regulated by Ets-TFs. This is concordance with data cdk-independent p27 transcriptional repression remains shown in Figure 5e lane 8, indicating that the p27-TGs unclear. Future work will be delineated to clarify these upregulated in p27À/À MEFs are also those regulated by possibilities. Ets-TFs. Collectively, these data indicate that p27 is The observation that in p27À/À cells a significant clearly a transcriptional repressor of genes under the number of p27-TGs are upregulated suggested that regulation of Ets-TFs. tumors with low levels of p27 could also upregulate a In contrast to what observed in p27À/À MEFs, a number of these genes. In fact, we observed that p27- significant number of p27-TGs repressed by p130/E2F4 TGs are overexpressed in a broad spectrum of cancers. complexes are upregulated in p27CKÀ MEFs (Figure 5e The high correlation between the p27-TGs overex- lane 6). These data indicate that the presence of p27 is pressed in a several sets of breast tumors and those needed for the transcriptional activation of these genes. overexpressed in p27À/À MEFs, also supports the idea Interestingly, the upregulation of these genes did not that the decrease of p27 in tumors can lead to the occur in p27WT MEFs, indicating that for the transcrip- increased expression of specific p27-TGs. For instance,

Oncogene Transcriptional regulatory role of p27Kip1 R Pippa et al 4217 two p27-TGs Aurka, encoding the protein kinase 0.1%. Triton X-100 and 1% bovine serum albumin in Aurora A, and Kif11, encoding a kinesin-related motor phosphate-buffered saline (PBS) solution. After three washes protein, are overexpressed in tumors and are considered with PBS, cells were incubated with primary antibody in as therapeutic targets for cancer, and a number of blocking buffer (1% bovine serum albumin in PBS solution) inhibitors are in clinical trials (Carter et al., 2006; Saijo for 1 h at 37 1C and then with secondary antibodies labeled et al., 2006; Malumbres and Barbacid, 2007; Huszar with Alexa fluor 488 (Invitrogen) (dilution 1:500). After a 10 min washing, incubation with 46-diamidino-2-phenyl indole et al., 2009; Lapenna and Giordano, 2009). The finding (dilution 1:10.000) was made. Stained coverslips were mounted À/À that both proteins are overexpressed in p27 cells using Mowiol (Calbiochem, Merck Chemicals Ltd., Nothing- supports a link between a decrease in p27, overexpres- ham, UK) and visualized using confocal microscope (TCS SL; sion of these proteins and cancer. This is also supported Leica microsystems, Wetzlar, Germany). by our observation of a clear correlation between p27 decrease and the increased expression of Aurka and Immunoprecipitation Fancd2 proteins in breast cancer. Of special relevance Serum starved NIH 3T3 or MEFs were scraped and washed are results showing that overexpression of p27-TGs is twice with PBS. Pellets were lysed in 1 ml of IP buffer (PBS associated with poor survival in a number of tumors. containing 0.5% Triton X-100, 1 mM EDTA, 100 mM sodium Overall, these data supports the link among p27- orthovanadate, 0.25 mM PMSF, complete protease inhibitor deficiency with overexpression of p27-TGs and poor mixture by Roche Applied Science (Penzberg, Germany), and survival in cancer. 1/25 volume of DNAse I (Sigma). After centrifugation, In light of the new data presented here, the role of samples were sonicated, and the protein was subsequently nuclear p27 as a tumor suppressor should be re- quantified using Lowry method (Lowry et al., 1951). The evaluated as it may also be linked to its new role as a supernatants (1 mg protein) were incubated for 1 h at 4 1C with transcriptional repressor through mechanisms that are 7 mg of anti-p27 antibody (C-19) crosslinked to protein A- Dynabeads following manufacturer’s instructions. The immu- cdk-independent. An additional role of p27 in the nocomplexes were extensively washed with IP buffer and organization of the p130/E2F4 repressor complexes immunoprecipitates were subsequently eluted with 0.1 M and in the activation of G1/S genes has also been taken citrate, pH 2.5, and boiled at 100 1C in Laemmli buffer for under consideration. western blot analysis. As a control, lysates were incubated with irrelevant crosslinked rabbit IgG.

Materials and methods Affinity chromatography Two affinity chromatography columns were generated with the Cell culture and transfection p27 fragments p27-NT (aa 1-110) and p27-CT (aa 110–198) NIH 3T3 cells were cultured in Dulbecco’s modified Eagle that were linked to sepharose 4B beads (Sigma-Aldrich), as Medium supplemented with 10% donor bovine serum. p27WT, previously described (Vera et al., 2007). Columns (1 ml) were p27CKÀ, p27D51, p27À/À and p130À/À MEFs were cultured in loaded with 50 mg of purified recombinant E2F4 or p130 Dulbecco’s modified Eagle Medium supplemented with 10% fragments (p130NT, aa 1-417 and p130CT, 967-1082). After fetal bovine serum. All cultures were maintained at 37 1C and extensive washing the bound proteins were eluted with 300 mM 5% CO2. Plasmids and shRNAs were transfected in NIH 3T3 of NaCl and eluates analyzed by western blot. cells using lipofectamine 2000 (Invitrogen, Karlsruhe, Ger- many) following manufacturer’s instructions. Chromatin immunoprecipitation ChIP assay was performed as previously described (Aguilera Antibodies et al., 2004). Briefly, chromatin from cross-linked cells was Polyclonal antibodies against p27 (C-19), p130 (C-20 and sonicated, incubated overnight with rabbit anti-p27, p130, KAB40), E2F4 (C-20), ETS1 (C-20), mSIN3A (K-20), E2F4, mSIN3A, HDAC1, 4, 5 and anti-IgG antibodies in HDAC1 (H-51), HDAC4 (H-92), HDAC5 (H-74), Heatr1 RIPA buffer, and precipitated with protein G/A-Sepharose. (sc-134699), cdk4 (C-20), cyclin D1 (DCS-6) or actin (C-2) Cross-linkage of the co-precipitated DNA-protein complexes were purchased from Santa Cruz Biotechnology (Santa Cruz, was reversed, and DNA was used as a template for qPCR. CA, USA). Monoclonal p27Kip1 and p130 antibodies were obtained from Transduction labs, BD Biosciences (Franklin Lakes, NJ, USA); monoclonal E2F4 (4E2F04) and polyclonal ChIP on CHIP p27 (ab7961) were purchased from Abcam (Cambridge, UK); Agilent G5590A mouse promoter microarray was used to and anti-Rabbit IgG whole molecule (R5506) was from Sigma- analyze p27-target genes in quiescent NIH 3T3 cells. Aldrich (Madrid, Spain). After scanning, raw data was extracted using Feature Extraction (Agilent Technologies, Basel, Switzerland). The Cell synchronization data analysis was performed using LIMMA package from R NIH 3T3 cells were maintained in serum-free medium for 48 h, of bioconductor, while the intrachip normalization was whereas MEFs were maintained for 72 h. Cells were then obtained with LOESS method. Only probes with a P-value activated by serum addition. At different time points after o0.02 in both experiments were considered as bound. serum re-stimulation, samples were taken and used for western blot and immunocytochemistry experiments. ChIP validation ChIP samples (2 ml) were used for the real time qPCR reaction Immunocytochemistry using Brilliant SybrGreen master mix from Agilent Technol- Cells were grown onto glass coverslips, fixed with 4% ogies. The primers used in the experiments are listed in the paraformaldehyde, then permeabilized and blocked with Supplementary Information (experimental procedures).

Oncogene Transcriptional regulatory role of p27Kip1 R Pippa et al 4218 Expression microarrays Functional enrichment analysis Cells were collected by trypsinization and then washed with Functional annotation of genes based on Gene Ontology PBS. Total RNA was extracted using the RNeasy Kit (Qiagen (GO) (Consortium, 2006) were extracted from Ensembl v.47. Iberia, Madrid, Spain). RNAs from quiescent p27WT, p27CKÀ, (Flicek et al., 2008). Enrichment of the genes from different gene p27D51 and p27À/À MEFs cells were analyzed using a ontology categories was done using GiTools (Perez-Llamas and Bioanalyzer (Agilent). Only samples with RIN number greater Lopez-Bigas, 2011). The probability of the overlap between two than nine and a ratio A260/280 more than two were further used gene sets was assessed comparing the observed number of in microarray analysis. Retrotranscription was made followed shared genes with the expected number estimated using the by labeling with biotin. Fragmented samples were then binomial distribution using the following formula, hybridized with Affymetrix 430 2.0 and 1.0 ST murine arrays, x  X½ Š n and analyzed with dChip Affymetrix software and tools. The PðXXxÞ¼ pið1 À pÞnÀi i normalization of the arrays was performed using the robust i¼0 multi-array average (RMA) method as implemented in the Affymetrix package of R bioconductor. Quality control n ¼ number of genes in the category in question (for example, measurements were obtained with ArrayQualityMetrics, while with a particular gene ontology term). the statistical analysis of significant hits with Limma (Smyth, x ¼ total number of genes in the category in question 2004). targeted by p27 or with mis-regulated expression in p27À/À (observed). p ¼ frequency of p27 target genes or with mis-regulated shRNA experiments À/À NIH 3T3 cells were transfected with MISSION shRNA expression in p27 (success rate). P control vector and a specific p27 shRNA with the following The resulting -value is corrected for multiple testing using sequence: Benjamini–Hochberg procedure (Benjamini and Hochberg, 50-CCGGGCGCAAGTGGAATTTCGATTTCTCGA- 1995). Results are displayed in color-coded matrices in which P GAAATCGAAATTCCACTTGCGCTTTTTG-30. -value is represented by a color-coded scale. Colors towards red P P Both were purchased from Sigma-Aldrich (Madrid, Spain). show significant -values (corrected -value o0.05), indicating P After 24 h of transfection, serum-free medium and puromycin strong deviation from random expectations. Grey indicates - were added to the cells during additional 24 h, until collection value is not significant after multiple testing corrections. of the samples. Total RNA was extracted using the RNeasy Promoter analysis Kit (Qiagen). RNA (1 mg) was reverse transcribed using High Capacity cDNA Archive Kit from Applied Biosystems (Life TF-binding sites in promoter sequences (1 kb upstream of the Technologies, Madrid, Spain) following the manufacturer’s transcription start site) were predicted with MatScan software et al. protocol. Then, 2 ml of cDNA were used for each real time (Blanco , 2006), using position weight matrices from et al. PCR using Brilliant SybrGreen qPCR master mix from TRANSFAC database (Matys , 2003). To test whether Stratagene. All the primers used are listed in the Supplemen- there was a significant deviation from random expectation for tary Information (experimental procedures). distribution of TF-binding sites, we compared the observed number of motifs in p27 target promoters with the expected according to binomial Poisson distribution. Luciferase assay Luciferase vectors were obtained by cloning specific regions IntOGen analysis from Aurka, Coq5 and Sfrs8 promoter sequences into a pGL2 To study the expression status of p27-TGs in cancer samples, basic vector. Primers for the selected genes were designed we used a collation of oncogenomic experiments from adding XhoI and SacI target sequences at 50 and 30, IntOGen (Gundem et al., 2010). We tested whether p27-TGs respectively. The primers used for this amplification were: were enriched by genes mis-regulated in those experiments Aurka forward 50-TTTGAGCTCCAATGCCGTTCTGTG using binomial test, as described above. ATTG-30, Aurka reverse 50-CCCCTCGAGCGAACGTGCCAGTGA Cancer survival analysis GTAAG-30; Overlapping between p27-TGs and genes overexpressed in Coq5 forward 50-TTTGAGCTCGCCAGCCTGGTCTATT human tumors with poor prognosis was done using the CAG-30, Oncomine gene expression database (Rhodes et al., 2004). Coq5 reverse 50-AAACTCGAGCTTCCGACACGGTCTC Studies of breast cancer, astrocytoma and lung adenocarcino- AAAC-30; and ma with very significant association (P-valueo10À4) were Sfrs8 forward 50-TTCGAGCTCCAACAAGATCCCCACT selected for further Kaplan–Meier analysis. Genes with TGTCTC-30, significant overexpression in the more malignant samples were Sfrs8 reverse 50-TTTCTCGAGCTTCACTCCGCCTTTCC selected to group the samples using hierarchical clustering TCTC-30. analysis. In all studies, patient samples were divided in two Amplification of promoter sequences was made by qPCR groups, and each group was analyzed in Kaplan–Meier curves using genomic DNA and cloning the PCR products into in order to assess their clinical behavior. pGEMt vectors (Promega, Madison, WI, USA). After transformation into JMJ109 bacteria and double digestion Tissue microarrays with the selected restriction enzymes, the promoter fragments Commercially available tissue microarrays (CC08-21, Cybrdi were finally inserted into the pGL2B vector. NIH 3T3 cells Inc., Rockville, MD, USA) containing normal breast tissue were co-transfected with CMV-bGal vector and a pGL2b and samples from 96 individual patients of breast ductal vector, empty or containing part of the promoters of Aurka, carcinoma of grade I, II, III and lymph node metastasis was Coq5 and Sfrs8 representing p27 targets and Noxa as a control de-waxed and treated by microwave boiling in citrate buffer, promoter. Luciferase assays (Luciferase Assay System; Pro- blocked by incubation in 10% non-immune horse serum and mega) were performed at 48 h after transfection, as previously incubated with the primary antibodies overnight at 4 1C. described (Aguilera et al., 2004). Antibodies used were a mAb against p27 (Clon 57; # 610242

Oncogene Transcriptional regulatory role of p27Kip1 R Pippa et al 4219 BD Biosciences) diluted 1/200, rabbit polyclonal Ab against Acknowledgements FancD2 (Abcam ab2187) diluted 1/500 and mAb against Aurka (Clon 35C1 Abcam ab13824) diluted 1/500. After We thank Dr Anxo Vidal for p27D51- MEF cells; Dr Antonio exhaustive washing in PBST, sections were incubated with Giordano for p130 pGEX plasmids; Dr Xavier Mayol for appropriate biotin-coupled secondary antibodies (all 1/1000 in GST-E2F4 vector; and Dr Alan Cassady for the pGL2b PBST) followed by avidin-peroxidase (ABC elite kit Vector, plasmid. This work was supported by the Ministerio de Vector Laboratories, Burlingame, CA, USA). Positive staining Ciencia y Tecnologı´a of Spain (grants SAF2006-05212 and was determined using diaminobenzidine as a substrate (DAB SAF2009-07769 to OB and SAF2009-06954 to NL-B; the kit Vector, Vector laboratories) following manufacturer’s Fondo de Investigacio´n Sanitaria (grant PI070778 to LE), the recommendations. Sections were then counterstained with Instituto de Salud Carlos III (RETICS RD06/0020/0010 to hematoxylin and mounted. Statistical analyses were carried OB, RD06/0020/0029 to JP and RD06/0020/0098 to A. Bigas), out with the SSPS program, version 11. 5. (SSPS, Chicago, IL, the Comunidad Autonoma de Madrid (Oncocycle Program USA). Frequencies were compared by the w2-contingency test. grant S2006/BIO-0232 to JP) and the Generalitat de Catalunya (grant SGR 09-1382 to OB). GG is supported by a fellowship from AGAUR of the Catalonian Government. AB is Conflict of interest supported by grants from the Association pour la recherche´ sur le Cancer, Ligue Nationale Contre le Cancer and Institue The authors declare no conflict of interest. National du Cancer.

References

Acosta JC, Ferrandiz N, Bretones G, Torrano V, Blanco R, Richard C Chu IM, Hengst L, Slingerland JM. (2008). The Cdk inhibitor p27 in et al. (2008). inhibits p27-induced erythroid differentiation human cancer: prognostic potential and relevance to anticancer of leukemia cells by repressing erythroid master genes therapy. Nat Rev Cancer 8: 253–267. without reversing p27-mediated cell cycle arrest. Mol Cell Biol 28: Cobrinik D, Lee MH, Hannon G, Mulligan G, Bronson RT, Dyson N 7286–7295. et al. (1996). Shared role of the pRB-related p130 and p107 proteins Aguilera C, Hoya-Arias R, Haegeman G, Espinosa L, Bigas A. (2004). in limb development. Genes Dev 10: 1633–1644. Recruitment of IkappaBalpha to the promoter is associated Fero ML, Rivkin M, Tasch M, Porter P, Carow CE, Firpo E et al. with transcriptional repression. Proc Natl Acad Sci USA 101: (1996). A syndrome of multiorgan hyperplasia with features of 16537–16542. gigantism, tumorigenesis, and female sterility in p27(Kip1)-deficient Baldassarre G, Belletti B, Nicoloso MS, Schiappacassi M, Vecchione mice. Cell 85: 733–744. A, Spessotto P et al. (2005). p27(Kip1)-stathmin interaction Flicek P, Aken BL, Beal K, Ballester B, Caccamo M, Chen Y et al. influences sarcoma cell migration and invasion. Cancer Cell 7: (2008). Ensembl 2008. Nucleic Acids Res 36: D707–D714. 51–63. Frescas D, Pagano M. (2008). Deregulated proteolysis by the F-box Benjamini Y, Hochberg Y. (1995). Controlling the false discovery rate: proteins SKP2 and beta-TrCP: tipping the scales of cancer. Nat Rev a practical and powerful approach to multiple testing. J R Stat Soc Cancer 8: 438–449. Series B (Methodol) 57: 289–300. Grimmler M, Wang Y, Mund T, Cilensek Z, Keidel EM, Waddell MB Besson A, Dowdy SF, Roberts JM. (2008). CDK inhibitors: cell cycle et al. (2007). Cdk-inhibitory activity and stability of p27Kip1 are regulators and beyond. Dev Cell 14: 159–169. directly regulated by oncogenic tyrosine kinases. Cell 128: 269–280. Besson A, Gurian-West M, Schmidt A, Hall A, Roberts JM. (2004). Gundem G, Perez-Llamas C, Jene-Sanz A, Kedzierska A, Islam A, p27Kip1 modulates cell migration through the regulation of RhoA Deu-Pons J et al. (2010). IntOGen: integration and data mining of activation. Genes Dev 18: 862–876. multidimensional oncogenomic data. Nat Methods 7: 92–93. Besson A, Hwang HC, Cicero S, Donovan SL, Gurian-West M, Gutie´rrez NC, Ocio EM, de Las Rivas J, Maiso P, Delgado M, Johnson D et al. (2007). Discovery of an oncogenic activity in Fermin˜a´nEet al. (2007). Gene expression profiling of B p27Kip1 that causes stem cell expansion and a multiple tumor lymphocytes and plasma cells from Waldenstro¨m’s macroglobuli- phenotype. Genes Dev 21: 1731–1746. nemia: comparison with expression patterns of the same cell Blain SW. (2008). Switching cyclin D-Cdk4 kinase activity on and off. counterparts from chronic lymphocytic leukemia, multiple myeloma Cell Cycle 7: 892–898. and normal individuals. Leukemia 21: 541–549. Blanco E, Messeguer X, Smith TF, Guigo R. (2006). Transcription Hollenhorst PC, Shah AA, Hopkins C, Graves BJ. (2007). Genome- factor map alignment of promoter regions. PLoS Comput Biol 2: wide analyses reveal properties of redundant and specific promoter e49. occupancy within the ETS gene family. Genes Dev 21: 1882–1894. Brune V, Tiacci E, Pfeil I, Do¨ring C, Eckerle S, van Noesel CJ et al. Huszar D, Theoclitou ME, Skolnik J, Herbst R. (2009). Kinesin motor (2008). Origin and pathogenesis of nodular lymphocyte-predomi- proteins as targets for cancer therapy. Cancer Metastasis Rev 28: nant Hodgkin lymphoma as revealed by global gene expression 197–208. analysis. J Exp Med 205: 2251–2268. James MK, Ray A, Leznova D, Blain SW. (2008). Differential Carter BZ, Mak DH, Shi Y, Schober WD, Wang RY, Konopleva M modification of p27Kip1 controls its cyclin D-cdk4 inhibitory et al. (2006). Regulation and targeting of Eg5, a mitotic motor activity. Mol Cell Biol 28: 498–510. protein in blast crisis CML: overcoming imatinib resistance. Cell Jones MH, Virtanen C, Honjoh D, Miyoshi T, Satoh Y, Okumura S Cycle 5: 2223–2229. et al. (2004). Two prognostically significant subtypes of high-grade Casey T, Bond J, Tighe S, Hunter T, Lintault L, Patel O et al. (2009). lung neuroendocrine tumours independent of small-cell and large- Molecular signatures suggest a major role for stromal cells in cell neuroendocrine carcinomas identified by gene expression development of invasive breast cancer. Breast Cancer Res Treat 114: profiles. Lancet 363: 775–781. 47–62. Kiyokawa H, Kineman RD, Manova-Todorova KO, Soares VC, Chu I, Sun J, Arnaout A, Kahn H, Hanna W, Narod S et al. (2007). Hoffman ES, Ono M et al. (1996). Enhanced growth of mice lacking p27 phosphorylation by Src regulates inhibition of cyclin E-Cdk2. the cyclin-dependent kinase inhibitor function of p27(Kip1). Cell 85: Cell 128: 281–294. 721–732.

Oncogene Transcriptional regulatory role of p27Kip1 R Pippa et al 4220 Lapenna S, Giordano A. (2009). Cell cycle kinases as therapeutic mavirus-negative head/neck and cervical cancers. Cancer Res 67: targets for cancer. Nat Rev Drug Discov 8: 547–566. 4605–4619. Lopez-Bigas N, Kisiel TA, Dewaal DC, Holmes KB, Volkert TL, Rayman JB, Takahashi Y, Indjeian VB, Dannenberg JH, Catchpole S, Gupta S et al. (2008). Genome-wide analysis of the H3K4 histone Watson RJ et al. (2002). E2F mediates cell cycle-dependent demethylase RBP2 reveals a transcriptional program controlling transcriptional repression in vivo by recruitment of an HDAC1/ differentiation. Mol Cell 31: 520–530. mSin3B corepressor complex. Genes Dev 16: 933–947. Lowry OH, Rosenbrough NJ, Farr AL, Randall RJ. (1951). Protein Rhodes DR, Yu J, Shanker K, Deshpande N, Varambally R, Ghosh D measurement with the folin-phenol reagent. J Biol Chem 193: et al. (2004). ONCOMINE: a cancer microarray database and 165–175. integrated data-mining platform. Neoplasia 6: 1–6. Macaluso M, Montanari M, Giordano A. (2006). Rb family proteins Richardson AL, Wang ZC, De Nicolo A, Lu X, Brown M, Miron A as modulators of gene expression and new aspects regarding the et al. (2006). X chromosomal abnormalities in basal-like human interaction with chromatin remodeling enzymes. Oncogene 25: breast cancer. Cancer Cell 9: 121–132. 5263–5267. Russo AA, Jeffrey PD, Patten AK, Massague J, Pavletich NP. (1996). Malumbres M, Barbacid M. (2005). Mammalian cyclin-dependent Crystal structure of the p27Kip1 cyclin-dependent-kinase inhibitor kinases. Trends Biochem Sci 30: 630–641. bound to the cyclin A-Cdk2 complex. Nature 382: 325–331. Malumbres M, Barbacid M. (2007). Cell cycle kinases in cancer. Curr Sabates-Bellver J, Van der Flier LG, de Palo M, Cattaneo E, Maake C, Opin Genet Dev 17: 60–65. Rehrauer H et al. (2007). Transcriptome profile of human colorectal Matys V, Fricke E, Geffers R, Gossling E, Haubrock M, Hehl R et al. adenomas. Mol Cancer Res 5: 1263–1275. (2003). TRANSFAC: transcriptional regulation, from patterns to Saijo T, Ishii G, Ochiai A, Yoh K, Goto K, Nagai K et al. (2006). Eg5 profiles. Nucleic Acids Res 31: 374–378. expression is closely correlated with the response of advanced non- McAllister SS, Becker-Hapak M, Pintucci G, Pagano M, Dowdy SF. small cell lung cancer to antimitotic agents combined with platinum (2003). Novel p27(kip1) C-terminal scatter domain mediates Rac- chemotherapy. Lung Cancer 54: 217–225. dependent cell migration independent of cell cycle arrest functions. Shedden K, Taylor JM, Enkemann SA, Tsao MS, Yeatman TJ, Gerald Mol Cell Biol 23: 216–228. WL et al. (2008). Gene expression-based survival prediction in lung Munoz-Alonso MJ, Acosta JC, Richard C, Delgado MD, Sedivy J, adenocarcinoma: a multi-site, blinded validation study. Nat Med 14: Leon J. (2005). p21Cip1 and p27Kip1 induce distinct cell cycle 822–827. effects and differentiation programs in myeloid leukemia cells. Sherr CJ, Roberts JM. (1999). CDK inhibitors: positive and negative J Biol Chem 280: 18120–18129. regulators of G1-phase progression. Genes Dev 13: 1501–1512. Nakayama K, Ishida N, Shirane M, Inomata A, Inoue T, Shishido N Shirane M, Harumiya Y, Ishida N, Hirai A, Miyamoto C, et al. (1996). Mice lacking p27(Kip1) display increased body size, Hatakeyama S et al. (1999). Down-regulation of p27(Kip1) by multiple organ hyperplasia, retinal dysplasia, and pituitary tumors. two mechanisms, ubiquitin-mediated degradation and proteolytic Cell 85: 707–720. processing. J Biol Chem 274: 13886–13893. Nguyen L, Besson A, Heng JI, Schuurmans C, Teboul L, Parras C Sicinski P, Zacharek S, Kim C. (2007). Duality of p27Kip1 function in et al. (2006). p27kip1 independently promotes neuronal differentia- tumorigenesis. Genes Dev 21: 1703–1706. tion and migration in the cerebral cortex. Genes Dev 20: 1511–1524. Slingerland J, Pagano M. (2000). Regulation of the cdk inhibitor p27 Pawitan Y, Bjohle J, Amler L, Borg AL, Egyhazi S, Hall P et al. and its deregulation in cancer. J Cell Physiol 183: 10–17. (2005). Gene expression profiling spares early breast cancer patients Smyth GK. (2004). Linear models and empirical Bayes methods for from adjuvant therapy: derived and validated in two population- assessing differential expression in microarray experiments. Stat based cohorts. Breast Cancer Res 7: R953–R964. Appl Genet Mol Biol 3: article 3. Perez-Llamas C, Lopez-Bigas N. (2011). Gitools: analysis and Soeiro I, Mohamedali A, Romanska HM, Lea NC, Child ES, visualisation of genomic data using interactive heat-maps. PLoS Glassford J et al. (2006). p27Kip1 and p130 cooperate to One 6: e19541. regulate hematopoietic cell proliferation in vivo. Mol Cell Biol 26: Phillips HS, Kharbanda S, Chen R, Forrest WF, Soriano RH, Wu TD 6170–6184. et al. (2006). Molecular subclasses of high-grade glioma predict Vera J, Estanyol JM, Canela N, Llorens F, Agell N, Itarte E et al. (2007). prognosis, delineate a pattern of disease progression, and resemble Proteomic analysis of SET-binding proteins. Proteomics 7: 578–587. stages in neurogenesis. Cancer Cell 9: 157–173. Viglietto G, Motti ML, Bruni P, Melillo RM, D’Alessio A, Califano D Piccaluga PP, Agostinelli C, Califano A, Rossi M, Basso K, Zupo S et al. (2002). Cytoplasmic relocalization and inhibition of the cyclin- et al. (2007). Gene expression analysis of peripheral T cell dependent kinase inhibitor p27(Kip1) by PKB/Akt-mediated lymphoma, unspecified, reveals distinct profiles and new potential phosphorylation in breast cancer. Nat Med 8: 1136–1144. therapeutic targets. J Clin Invest 117: 823–834. Yeh N, Miller JP, Gaur T, Capellini TD, Nikolich-Zugich J, de la HC Pyeon D, Newton MA, Lambert PF, den Boon JA, Sengupta S, et al. (2007). Cooperation between p27 and p107 during endochon- Marsit CJ et al. (2007). Fundamental differences in cell cycle dral ossification suggests a genetic pathway controlled by p27 and deregulation in human papillomavirus-positive and human papillo- p130. Mol Cell Biol 27: 5161–5171.

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

Oncogene