Prox1 Suppresses the Proliferation of Neuroblastoma Cells Via a Dual Action in P27-Kip1 and Cdc25a

ORIGINAL ARTICLE Prox1 suppresses the proliferation of neuroblastoma cells via a dual action in p27-Kip1 and Cdc25A

IP Foskolou, D Stellas, I Rozani, MD Lavigne and PK Politis

Neuroblastoma is a pediatric tumor that originates from precursor cells of the sympathetic nervous system with less than 40% long- term survival in children diagnosed with high-risk disease. These clinical observations underscore the need for novel insights in the mechanisms of malignant transformation and progression. Accordingly, it was recently reported that Prox1, a homeobox transcription regulator, is expressed in higher levels in human neuroblastoma with favorable prognosis. Consistently, we have recently shown that Prox1 exerts a strong antiproliferative effect on neural precursor cells during embryonic development. Thus, Prox1 is a candidate gene with a critical role in suppressing malignant neuroblastoma transformation. Here, we provide evidence that Prox1 strongly suppresses the proliferation of mouse and human neuroblastoma cell lines and blocks the growth of neuroblastoma tumors in SCID mice. Conversely, short hairpin RNA (shRNA) -mediated knockdown of basal Prox1 expression significantly induces proliferation, genomic instability and the ability of neuroblastoma cells to form tumors. Mechanistically, analysis of an inducible Prox1-overexpressing Neuro2A cell line indicates that Prox1 is sufficient to suppress CyclinD1, CyclinA and CyclinB1, consistent with a role in cell cycle arrest. Surprisingly, Prox1 strongly induces CyclinE1 expression in the same system despite its action on blocking cell cycle progression, which could account for the context dependent oncogenic function of Prox1. Most importantly, Prox1 was sufficient to decrease Cdc25A and induce p27-Kip1, but not p21-Cip1 or p53. By alleviating the Prox1 action in Cdc25A and p27-Kip1 expression, we were able to rescue its effect on cell cycle arrest. Together these data suggest that Prox1 negatively regulates neuroblastoma carcinogenesis through suppression of Cdc25A and induction of p27-Kip1 to counteract CyclinE1 overexpression and block cell cycle progression. Furthermore, these observations render Prox1 a candidate target for the treatment of neuroblastoma tumors.

Oncogene (2013) 32, 947–960; doi:10.1038/onc.2012.129; published online 16 April 2012 Keywords: prospero; cyclinE1; G0/G1-arrest; cell cycle; SH-SY5Y; tet-on

INTRODUCTION of proliferation and differentiation in neural precursor cells.5 In Neuroblastoma is a malignancy of early childhood that arises from agreement, recent data suggest that Notch1 expression predicts the developing autonomic nervous system.1 The clinical behavior an unfavorable prognosis for patients with neuroblastoma and is variable, ranging from spontaneous regression to highly could serve as a therapeutic target.6 Therefore, Prox1 may be aggressive metastatic disease with a poor overall survival rate. It involved in the malignant transformation, progression and/or is the most frequently diagnosed neoplasm during infancy and regression of neuroblastoma through Notch signaling regulation less than 40% of children with high-risk neuroblastoma are likely and/or additional genes and pathways. to achieve long-term cure, despite dramatic escalations in the Multiple lines of evidence suggest important roles for Prox1 in intensity of therapy provided.2 Future therapies will focus different aspects of embryonic development and morphogenesis, increasingly on the genes and biological pathways that while mouse embryos deﬁcient in Prox1 die at E14.5. Prox1 has contribute to this malignancy. been previously shown to have essential roles during lymphatic, Towards this aim, it was recently reported that expression of hepatocyte, pancreatic, heart, lens, retinal and spinal cord Prox1, a homeobox transcription regulator, in human neuroblas- development.5,7–10 Prox1 has also been implicated in both toma tumors is correlated with the stage 4S, which has favorable progression and suppression of malignancies. A function for prognosis with high incidence of spontaneous regression. More- Prox1 as tumor suppressor has been described previously in other over, it was shown that Prox1 is highly expressed in sympathetic systems. Decreased expression levels have been found in neurons during early stages of development in chick, mouse and hepatocellular and pancreatic carcinoma,11,12 which may be due human peripheral nervous system and its expression levels are to epigenetic silencing. Hypermethylation of the Prox1 gene has reduced in highly proliferative human neuroblastoma cell lines.3,4 been found in breast cancer and lymphomas and in brain These observations suggest that Prox1 might have a tumor metastases of breast cancer.13,14 Moreover, mutations of Prox1 suppressor function in neuroblastoma cancer. Accordingly, we and loss of heterozygosity have been reported in various have recently reported that Prox1 negatively regulates the cancers.13,15,16 Although most studies point to a tumor proliferation of neural precursor cells during embryonic suppressor function of Prox1, a recent study reports that it development. Furthermore, we showed that a cross-inhibitory enhances colorectal cancer progression.17 These data suggest that interaction between Prox1 and Notch1 is involved in the regulation Prox1 may function in a context dependent manner, being a

Center for Basic Research, Biomedical Research Foundation of the Academy of Athens, Athens, Greece. Correspondence: Dr PK Politis, Center for Basic Research, Biomedical Research Foundation of the Academy of Athens, 4 Soranou Efesiou, Athens 11527, Greece. E-mail: [email protected] Received 1 November 2011; revised 28 February 2012; accepted 7 March 2012; published online 16 April 2012 Prox1 blocks cell cycle progression in neuroblastoma cells IP Foskolou et al 948 tumor suppressor in certain cells, while in other cell types could be RESULTS related to tumor progression. This dual function may reside in the Prox1 blocks proliferation of neuroblastoma cancer cells fact that Prox1 affects differentially the expression of genes that To evaluate whether the observed correlation between Prox1 promote or inhibit proliferation and cell cycle progression. expression and suppression of neuroblastoma progression has Here, we provide functional evidence that Prox1 is indeed functional importance, we used neuroblastoma cell lines as a involved in the suppression of neuroblastoma cell growth both model system to analyze proliferation. We ﬁrst examined the in vitro and in vivo. Mechanistically, activated Notch1 signaling effect of mouse or human Prox1 overexpression in Neuro2A (N2A) cannot rescue the negative effect of Prox1 on neuroblastoma mouse neuroblastoma cell line. BrdU incorporation assays proliferation, suggesting an alternative mode of action. Accord- revealed a strong block in proliferation after mouse or human ingly, we showed that Prox1 acts in basic components of cell cycle Prox1 overexpression (Figures 1a and b and Supplementary Figure machinery to regulate cellular proliferation in neuroblastoma cells. S1a). In addition, immunostainings with phosphorylated-histo- Collectively, these observations indicate that Prox1 negatively neH3 (pH3) revealed a strong reduction in Prox1 þ N2A cells regulates tumorigenic properties of neuroblastoma cells, may be undergoing mitosis (Figures 1d and e and Supplementary Figure implicated in suppressing neuroblastoma carcinogenesis, and S1b). To exclude the possibility of non-cell autonomous effects, we could be utilized for its treatment. also measured the proliferation rate in the nontransfected cells. In

Figure 1. Prox1 has a strong antiproliferative effect on mouse and human neuroblastoma cell lines. (a) Prox1 or GFP transfected N2A cells were pulsed for 2 h with BrdU and then labeled for BrdU and 4,6-diamidino-2-phenylindole (DAPI). Scale bar: 25 mm. (b) Quantification of the transgene-positive/BrdU-positive cells. (GFP: 43.6±4.26%, Prox1: 1.86±0.59%; Po0.01) (c) Quantification of the transgene negative/BrdU- positive cells (GFP: 40.56±3.01%, Prox1: 39.13±1.66%; P40.1). (d) Prox1- or GFP-transfected N2A cells were labeled for pH3 and DAPI. Scale bar: 25 mm. (e) Quantification of the transgene-positive/pH3-positive cells (GFP: 11.52±1.2%, Prox1: 1.09±0.25%; Po0.05). (f) Quantification of the transgene-negative/pH3-positive cells (GFP: 11.66±0.47%, Prox1: 11.47±1.2%; P40.1). (g, h) Quantification of the transgene-positive/ BrdU-positive cells in SH-SY5Y and Kelly human neuroblastoma cell lines, respectively, (GFP: 23.78±1.68%, Prox1: 3.89±0.59%; Po0.01 in SH-SY5Y cells; GFP: 51.37±5.29%; Prox1: 20.5±0.86%; Po0.05 in Kelly cells).

Oncogene (2013) 947 – 960 & 2013 Macmillan Publishers Limited Prox1 blocks cell cycle progression in neuroblastoma cells IP Foskolou et al 949 all cases, the untransfected cells in Prox1 experiments showed observations with MTT assay and PCNA immunostainings (Figures proliferation rates similar to green fluorescent protein (GFP) 3k–m). Collectively, these observations show that Prox1 negatively experiments (Figures 1c and f), further confirming that the defect affects proliferative and tumor forming abilities of neuroblastoma in proliferation is because of Prox1 overexpression. Prox1 was also cells in vitro with no indication of increased cell death. sufficient to suppress proliferation in human neuroblastoma cell lines, including SH-SY5Y and Kelly (Figures 1g and h). We next asked whether the lower proliferative rate of Prox1 transfected shRNA-mediated knockdown of Prox1 enhances proliferation of cells is owing to cell death. No evidence of induction of apoptosis neuroblastoma cells was observed in Prox1 transfected cells compared with GFP or RFP To test whether the low, basal expression levels of Prox1 in as documented by TUNEL assay and PE/AnnexinV fluorescence- neuroblastoma cell lines3 are involved in regulating proliferation activated cell sorting analysis (FACS) (Figure 2). To further validate of these cells, we constructed a set of stable cell lines, expressing the Prox1-mediated impairment of cell proliferation, we trans- different shRNAs targeting the mouse Prox1 homolog. To this end, fected N2A cells with either GFP or Prox1-GFP and we then we utilized five different shRNAs constructs for Prox1 and one analyzed the clonal formation of fluorescent cells (Supplementary control scrambled construct based on the lentiviral TRC library Figure S2). Although GFP þ clones tend to form large colonies (Sigma, St Louis, MO, USA). In particular, we infected WT N2A cells with many cells few days after plating, Prox1-GFP þ clones stayed with the indicated lentiviruses and subjected the infected cells in single cells. Most of them stayed quiescent without dying for at puromycin selection to construct six distinct stable cell lines. One least 7 days after plating (Supplementary Figure S2 and data not of them, designated shRNA-55 (hereafter referred as shProx1), was shown), whereas untransfected cells in the same well formed large capable to strongly downregulate the endogenous Prox1 mRNA colonies (that is, Supplementary Figure S2a, Day4, phase contrast). levels (Figure 4a). We further confirmed the ability of this vector to To further study the involvement of Prox1 in neuroblastoma cell downregulate Prox1 expression by constructing a set of stable cell growth both in vitro and in vivo, we constructed a set of inducible lines based on Prox1-Tet, able to induce Prox1 and stably express Tet-On N2A stable cell lines capable to efficiently induce Prox1 or this shRNA as indicated in Figure 4b. Thus, shProx1 was sufficient GFP expression (Figures 3a and b). First, we confirmed that Prox1 to block the Dox-mediated induction of Prox1 in the Prox1-Tet cell induction after doxycycline (Dox) administration strongly inhibits background, as compared with the scrambled or another shRNA proliferation of these cells (Prox1-Tet) as compared with vector (shRNA-57) (Figures 4c and d). Remarkably, the shProx1 noninduced or GFP-induced (GFP-Tet) cells (Figures 3c–j), with stable cell line, with much reduced levels of endogenous Prox1 no induction of apoptosis (data not shown). We verified these (Figure 4a), showed significantly enhanced proliferative rates as

Figure 2. Prox1-mediated impairment of proliferation is not accompanied by induction of apoptosis. (a) Prox1- or RFP-transfected N2A cells were assayed with TUNEL reaction (green) and counterstained with 4,6-diamidino-2-phenylindole (DAPI). Scale bar: 25 mm. (b) Quantiﬁcation of apoptosis in transfected Prox1 or RFP cells. Results are expressed as the numbers of Prox1 or RFP-positive/TUNEL-positive cells (RFP: 1.68±0.52%, Prox1: 0.64±0.19%; Po0.05). (c) Prox1-GFP- or GFP-transfected N2A cells were labeled with PE Annexin V and then was determined the percentage of cells within the population that were undergoing apoptosis with ﬂuorescence-activated cell sorting (FACS) analysis (GFP: 4.3%, Prox1-GFP: 2.6%).

& 2013 Macmillan Publishers Limited Oncogene (2013) 947 – 960 Prox1 blocks cell cycle progression in neuroblastoma cells IP Foskolou et al 950

Figure 3. Construction and analysis of Tet-On inducible overexpression systems for Prox1 and GFP in N2A cell line. (a, b) Western blot analysis of protein extracts from Prox1-Tet and GFP-Tet cells, respectively, in the presence or absence of Dox. (c) Prox1-Tet cells were pulsed for 2 h with BrdU and then labeled for BrdU and 4,6-diamidino-2-phenylindole (DAPI) in the presence or absence of Dox. Prox1 was detected with an anti- flag antibody. (d) Quantification of the data ( À Dox: 9.7±0.72%, þ Dox: 0.63±0.25%; Po0.01). (e) GFP-Tet cells were pulsed for 2 h with BrdU and then labeled for BrdU and DAPI in the presence or absence of Dox. (f) Quantification of the data ( À Dox: 17. 75±0.49%, þ Dox: 18.6±0.38%; P40.1). (g) Prox1-Tet cells were labeled for pH3. (h) Quantification of the data ( À Dox: 7.47±0.25%, þ Dox: 1.03±0.31%; Po0.01). (i) GFP-Tet cells were labeled for pH3. (j) Quantification of the data ( À Dox: 15.1±0.99%, þ Dox: 15.85±0.78%; P40.1). (k) GFP-Tet and Prox1-Tet cells were tested by MTT assay in the presence or absence of Dox for 24 and 48 h. Note the strong inhibition of cell growth in the Prox1-Tet/ þ Dox. (l) GFP-Tet cells and Prox1-Tet cells in the presence of Dox were labeled with PCNA (red) and counterstained with DAPI. (m) Quantification of the data (GFP/ þ Dox: 56.56±3.46%,Prox1/ þ Dox: 18.76±0.92%; Po0.01). For all panels scale bars: 50 mm.

Oncogene (2013) 947 – 960 & 2013 Macmillan Publishers Limited Prox1 blocks cell cycle progression in neuroblastoma cells IP Foskolou et al 951

Figure 4. shRNA-mediated knockdown of basal Prox1 expression enhances cell proliferation in N2A. (a) Relative expression levels of Prox1 mRNA in N2A stable cell lines expressing different shRNAs. In particular, N2A cells were first infected with five different lentiviruses expressing distinct shRNAs targeting Prox1 (sh54-sh58) and one control scrambled shRNA (SCR), and were then selected with puromycin to construct six different stable cell lines. RNA was isolated from these cells and assayed for Prox1 mRNA with quantitative real-time RT–PCR. (b) Schematic representation of the construction of three stable cell lines able to induce Prox1 in the presence of Dox (Prox1-Tet) and stably express shSCR or shRNA-55 (shProx1) or shRNA-57. (c) Western blot analysis of protein extracts from these cells in the presence of Dox. (d) Detection of Prox1 transgene (green) by immunostaining in N2A stable cell lines as indicated. (e–h) shProx1 cell line showed significantly enhanced proliferative rates as compared with shSCR control cell line. (e) shProx1 and shSCR cells were pulsed for 2 h with BrdU and then labeled for BrdU and 4,6- diamidino-2-phenylindole (DAPI). (f) Quantification of the data (shSCR: 39.12 ±2.09, shProx1: 54.58±1.75; Po0.01). (g) shProx1 and shSCR cells were labeled for pH3 and counterstained with DAPI. (h) Quantification of the data (shSCR: 13.7±2.77, shProx1: 21.02±0.93; Po0.05). (i) Representative images of cells with larger nuclei in shProx1 cells compared with shSCR cells, an indication of genomic instability. Arrows indicate some of the large nuclei. (j) Quantification of the data. For all panels scale bars: 50 mm.

& 2013 Macmillan Publishers Limited Oncogene (2013) 947 – 960 Prox1 blocks cell cycle progression in neuroblastoma cells IP Foskolou et al 952 compared with shSCR (shScramble) control cell line (Figures 4e–h). The Prox1 effect on neuroblastoma proliferation is independent of In agreement, this cell line exhibited increased levels of genomic its action on Notch signaling instability, manifested by higher percentage of cells with large We have previously shown that Prox1 inhibits Notch1 expression nuclei (Figures 4i and j), indicative of aberrant control of at the transcriptional level.5 Notch signaling has a critical role in proliferation. To exclude off-target effects and assess the neuroblastoma pathogenesis, particularly related to its key specificity of Prox1-shRNA phenotypes, we repeated the same function in neural development.18 Activated Notch1 signaling set of experiments with the shRNA-56 cell line (Supplementary inhibits neuronal differentiation and maintains precursors of the Figure S3). This cell line showed enhanced proliferative capacity autonomic and central nervous system in a proliferative state.19 similar to shRNA-55, further confirming the effect of Prox1 knockdown on proliferation and genomic instability. Together these data suggest that Prox1 is sufficient and necessary for the proper regulation of proliferation in neuroblastoma cells.

Prox1 strongly suppresses neuroblastoma tumor growth in vivo To examine whether the Prox1-mediated inhibition of cell proliferation is recapitulated under more physiological conditions, Prox1-Tet and GFP-Tet clones were transplanted subcutaneous in SCID mice. We transplanted Prox1-Tet (right side) and GFP-Tet (left side) clones as pairs in the same animals, as indicated in Figure 5a. The effect of Prox1 induction in vivo on tumor growth was determined by measuring the volume of tumors every 5–8 days (Figure 5e). Remarkably, Prox1 was sufﬁcient to totally block the growth of tumors in SCID mice (Figures 5b–e). In particular, out of 10 transplantations for the GFP-Tet clone nine tumors were grown after 21 days with an average volume 2.84±1.08 cm3, whereas Prox1 gave in the same mice only three detectable tumors with much more reduced volume, 0.014±0.02 cm3 (Figure 5e). Detailed immunohistochemical analysis showed that the Prox1-Tet tumors were formed mainly because of escaper cells from Prox1 induction after Dox administration (Figures 5f–i). In the Prox1-Tet tumor tissue, cells that overexpress Prox1 showed reduced percentage of pH3 þ index (Figures 5j and k), with no indication of apoptosis (Supplementary Figure S4), whereas cells that escaped Prox1 induction had pH3 þ index similar to that of the GFP þ cells in GFP-Tet tumors (Figure 5j and data not shown). Thus, as was the case in vitro, the reduced N2A tumor growth rate was owing to reduced cell proliferation. Conversely, a signiﬁcant increase was observed in the volume of shProx1 tumors, compared with the control shSCR tumors, as early as 14 days after transplantation of both cell lines (Figures 5l and m). Taken together, our results demonstrate that Prox1 suppresses the oncogenic properties of neuroblastoma cells in vivo.

Figure 5. Allotransplantations in SCID mice. (a) Schematic representation of the injections of the Prox1-Tet (right side) and GFP-Tet (left side) clones as pairs in the same animals. The administration of Dox to these mice was initiated two days before allotransplantations via drinking water. (b) Representative pictures of the SCID mice 21 days after the transplantations. Out of 10 transplantations for the GFP-Tet clone, nine tumors were grown (average volume: 2.8±1.08 cm3), whereas Prox1 gave in the same mice only three detectable tumors (average volume: 0.014±0.02 cm3); Po0.01. (c) Representative picture of a mouse indicating the difference between the two tumors. (d) Representative picture of the tumors that were grown by GFP þ cells in comparison with Prox1 þ cells, respectively. (e) Quantification of the volume of tumors every 5–8 days. (f–g’) Immunohistochemical analysis of the tumors with hematoxylin and eosin (H&E). (h–i’) Immunostainings in the tumors with GFP or Prox1 (green) and 4,6-diamidino-2-phenylindole (DAPI, blue). Scale bar for (h) and (i): 250 mm; Scale bar for (h’) and (i’): 50 mm. (j) The GFP or Prox1 (green) tumors were labeled for pH3 (red). (k) Quantification of the data (GFP: 16.13±0.15, Prox1: 1.77±0.67; Po0.01). Scale bar for j:50mm. (l) Schematic representation of the injections of the shProx1 (right side) and shSCR (left side) clones as pairs in the same animals. (m) Quantification of the volume of tumors 14 days after transplantations (shProx1 average volume: 2.80±1.05 cm3, shSCR average volume: 1.37±0.86 cm3; Po0.01).

Oncogene (2013) 947 – 960 & 2013 Macmillan Publishers Limited Prox1 blocks cell cycle progression in neuroblastoma cells IP Foskolou et al 953 These data raised the possibility that Prox1 may act through blot analysis for p27-Kip1 and p21-Cip1. Interestingly, Prox1 was Notch1 signaling inhibition to achieve its antiproliferative action. sufficient to induce p27-Kip1 but not p21-Cip1 or its upstream Thus, co-expression of constitutively active form of Notch1 (NICD) activator p53 (Figures 8a,b and d). Moreover, we confirmed the was not sufficient to rescue the antiproliferative effect of Prox1 on Prox1 effect on p27-Kip1 expression in human SH-SY5Y cells neuroblastoma cells (Figures 6a–e). These results suggest that Prox1 induces cell cycle exit through the regulation of additional genes and/or pathways other than Notch1. Moreover, we have previously shown that pharmacological inhibition of Notch signaling induces the expression of Prox1 in neural precursor cells.5 However, in neuroblastoma cells even high doses of DAPT inhibitor cannot induce the expression of endogenous Prox1 (Supplementary Figure S5). Accordingly, DAPT-mediated inhibition of Notch signaling was able to only partially affect proliferation of N2A cells (Figure 6f and data not shown), consistent with the inability of DAPT inhibitor to induce Prox1. Therefore, the mild effect of Notch inhibition on N2A proliferation is also independent of Prox1 function.

Prox1 blocks cell cycle progression in G0/G1 phase To further investigate the mechanism of Prox1 action, we analyzed the cell cycle properties of Prox1-Tet cells after Dox induction. FACS analysis carried out 48 h after Dox induction indicated a cell cycle exit in G0/arrest in G1 phase as compared with the GFP-Tet cells (Figures 7a and b), consistent with the observations that Prox1-Tet do not enter S phase and do not accumulate around mitosis, as shown by BrdU and pH3 assays, respectively (Figures 3c–j). In agreement, Prox1 induction was sufficient to reduce the levels of CyclinA, CyclinB1 and CyclinD1 (Figure 7c and Supplementary Figure S6). Especially for the G1-specific CyclinD1, we confirmed this finding by measuring the index of cells that co- express CyclinD1 and Prox1 or GFP in Prox1-Tet versus GFP-Tet cells, respectively (Figures 7e and f). Moreover, Prox1 affects CyclinD1 in a similar manner in human neuroblastoma SH-SY5Y cells (Supplementary Figures S7a and S7b). Strikingly, despite the effect on cell cycle arrest, Prox1 was also able to enhance the expression of CyclinE1 (Figures 7c,d,g and h). In these cultures, endogenous CyclinE1 is upregulated upon induction of Prox1 at the protein level (Figures 7c and d) and a concomitant induction in the number of CyclinE1 þ cells is also observed in N2A cells (Figures 7g and h), as well as in human SH-SY5Y cells (Supplementary Figure S7c–7f). Similar to previous reports,20 the Prox1-mediated induction of CyclinE1 was at the transcription level (Figures 7i and j). To further confirm that Prox1 differentially affects the expression of G1 Cyclins, we analyzed the expression of CyclinD1 and E1 in sections from Prox1-Tet and GFP-Tet tumors. Similar to the in vitro data, immunohistochemical analysis in tumors indicates that Prox1 suppresses the expression of CyclinD1 and induces the expression of CyclinE1 (Figures 7k–n). We then asked whether Prox1 affects the expression of cdk inhibitors that act in G1 phase. To this end, we performed western

Figure 6. The antiproliferative effect of Prox1 is not mediated through the Notch1 pathway. (a) Co-expression of Prox1 (green) with a constitutively active form of Notch1 (NICD) (red) in N2A cells. The transgenes were detected with anti-Flag and anti-Myc antibodies, as indicated. Note that all Prox1 þ cells are also NICD þ . (b) Immunostaining in N2A cells that were co-transfected with Prox1 (green) and NICD, pulsed for 2 h with BrdU and then labeled for BrdU (red) and 4,6-diamidino-2-phenylindole (DAPI) (blue). (c) Quantiﬁcation of the data. (d) Immunostaining in N2A cells that were co-transfected with Prox1 (green) and NICD, and then labeled for pH3 (red) and counterstained with DAPI (blue). (e) Quantiﬁcation of the data. (f) N2A cells were either treated with 0, 10, 25 or 50 mM of DAPT for 3 days and then were tested by MTT assay. The treatment of the N2A cells with DAPT, only partially affect the proliferative capacity of the cells. For all panels scale bars: 50 mm.

& 2013 Macmillan Publishers Limited Oncogene (2013) 947 – 960 Prox1 blocks cell cycle progression in neuroblastoma cells IP Foskolou et al 954 (Supplementary Figures S8a and b). To determine whether Prox1 In addition, given that Prospero, the Drosophila homolog of affects the expression of p27-Kip1 at the transcriptional level, we Prox1, suppresses the expression of Drosophila homologue for measured the p27-Kip1 mRNA levels in Prox1-Tet and GFP-Tet Cdc25A (String),21,22 we sought to test whether Prox1 also affects cells. The Prox1-Tet cells showed signiﬁcantly enhanced mRNA Cdc25A expression in neuroblastoma cells. Accordingly, a levels of p27-Kip1, indicative of an action at the level of concomitant reduction in expression of the Cdc25A phosphatase transcriptional regulation (Figure 8e). was also evident after Prox1 expression in our inducible N2A cell

Oncogene (2013) 947 – 960 & 2013 Macmillan Publishers Limited Prox1 blocks cell cycle progression in neuroblastoma cells IP Foskolou et al 955 line (Figures 8a and c), as well as in SH-SY5Y cells (Supplementary However, additional mechanisms should operate, as Cdc25A Figure S8c–f). Similar to p27-Kip1, real time RT–PCR and confers only a partial rescue of the Prox1 effect. transcriptional assays in N2A cells suggest that Prox1 is sufficient Accordingly, we then tested whether the ability of Prox1 to to repress Cdc25A expression at the transcriptional level (Figures induce p27-Kip1 is also involved in this function. First, p27-kip1 8f and g). Considering that the activity of Cdc25A is required in overexpression was sufficient to recapitulate the Prox1 effect on mammalian cells for S phase progression,23 it could be one of the cell proliferation (Supplementary Figure S11). Second, to further putative targets of Prox1 antiproliferative function. examine this scenario, we utilized a previously tested shRNA To further investigate whether the Prox1-mediated regulation construct that targets and efficiently downregulates the mouse of p27-Kip1 (encoded by Cdkn1b gene) and Cdc25A is direct, we homolog for p27-Kip1 (confirmed in Supplementary Figure S12).25 performed chromatin immunoprecipitation experiments. In parti- We then transiently transfected N2A cells with the indicated cular, by using a series of PCR primer pairs we screened the constructs and employed BrdU incorporation and pH3 assays to corresponding gene promoters for Prox1-binding events (Figures measure proliferation (Figures 10e–h). Hence, shRNA-mediated 9a and c). Thus, we were able to identify genomic loci close to the knockdown of endogenous p27-Kip1 in N2A cells can significantly transcription start sites of Cdkn1b and Cdc25A genes that were rescue the negative effect of Prox1 on cell proliferation, by 48.36 specifically enriched in our chromatin immunoprecipitation assays and 58.87%, respectively. To further confirm these observations, for Prox1-Tet cells compared with GFP-Tet cells or compared with we transfected the Prox1-Tet and GFP-Tet cells with the shRNA other loci in the same genomic region (Figures 9b and d). These construct for p27-Kip1. Similar to the transiently transfected N2A data suggest that Prox1-mediated transcriptional regulation of cells, shRNA-mediated knockdown of p27-Kip1 in the Prox1-Tet Cdkn1b and Cdc25A genes may be through direct interactions with cells was again sufficient to partially rescue Prox1 effect on their promoters. proliferation (Supplementary Figure S13). Next, to test whether Taken together these observations suggest that Prox1 may act simultaneous misexpression of Cdc25A and ablation of p27-Kip1 on proliferation of neuroblastoma cells through a combined effect can fully rescue the Prox1 effect, we did triple transfection on various components of the cell cycle control machinery, experiments with the shRNA for p27-Kip1, Cdc25A and Prox1 including Cyclins, p27-Kip1 and Cdc25A. expression vectors. Interestingly, by simultaneously alleviating the action of Prox1 on the expression levels of Cdc25A and p27-Kip1, we were able to almost fully rescue the Prox1-mediated Cdc25A overexpression and p27-Kip1 knockdown rescue the impairment of cell proliferation (Figures 10i–l). antiproliferative effect of Prox1 Finally, by using a series of deletion mutants for the Prox1 To assess whether the observed correlations between Prox1 and protein in proliferation assays, we showed that various domains the components of cell cycle control machinery have functional are differentially required for mediating its antiproliferative effect importance, we performed a series of overexpression and (Supplementary Figure S14). Almost all domains showed some knockdown experiments in N2A cells in an attempt to recapitulate degree of rescue ability, in agreement with a combinatorial action or rescue the Prox1-mediated effect on cell proliferation. We first of Prox1 in various components of the cell cycle regulatory examined the effect of CyclinE1 overexpression on N2A cell machinery. In all cases, deletion of the DNA binding domain from proliferation (Supplementary Figure S9a). BrdU incorporation the carboxy terminal of Prox1 protein was sufficient to fully rescue analysis and pH3 immunostainings 48 h after transfection revealed the antiproliferative effect, consistent with the observation that a significant induction in proliferation (Supplementary Figures S9b Prox1 action on Cdc25A and p27-Kip1 occurs primarily at the and e). Therefore, both indices show that CyclinE1 overexpression transcriptional level. induces the proliferation of N2A cells, in clear contrast to Prox1 Taken together these observations indicate that Prox1 nega- misexpression, indicating that CyclinE1 cannot recapitulate the tively regulates neuroblastoma carcinogenesis through its ability Prox1 effect on growth of neuroblastoma cells. to suppress Cdc25A and induce p27-Kip1 to counteract CyclinE1 We next asked whether the Prox1-mediated suppression of overexpression and block cell cycle progression (Figure 10m). Cdc25A could explain its strong antiproliferative action. To Furthermore, these observations render Prox1 a candidate target this end, we co-expressed Cdc25A24 together with Prox1 for the treatment of neuroblastoma tumors. (Supplementary Figure S10) and measured proliferation by BrdU and pH3 48 h after transfection (Figures 10a–d). Interestingly, Cdc25A co-expression was able to partially rescue the effect of DISCUSSION Prox1, by 32.1 and 33.3%, respectively. These observations Recent data suggest that Prox1 transcription regulator is involved suggest that the Prox1-mediated downregulation of Cdc25A is in the early events of neuroblastoma carcinogenesis, thus being significantly involved in the antiproliferative function of Prox1. potential target for cancer therapy.3,4,26 These correlative data

Figure 7. Prox1 blocks cell cycle progression in G0/G1 phase. (a) Fluorescence-activated cell sorting (FACS) analysis in GFP-Tet and Prox1-Tet cells after 48 h Dox induction. (b) Quantification of the data. (c) Western blot analysis of protein extracts from GFP-Tet and Prox1-Tet cells in the presence of Dox for CyclinA, CyclinB1, CyclinD1, CyclinE1 and actin as indicated in 48 and 72 h. Note that loading controls (Actin) for Prox1, GFP and CyclinD1 western blots are shown in Supplementary Figure S6. (d) Quantification of the CyclinE1 protein blots. For both cases: Po0.01. (e) Immunostainings in GFP-Tet and Prox1-Tet cells for GFP or Prox1 (green), CyclinD1(red) and counterstained for 4,6-diamidino-2- phenylindole (DAPI). (f) Quantification of the data (GFP: 61.23±4.29%, Prox1: 22.17±2.01%; Po0.01). (g) Immunostaining in GFP-Tet and Prox1-Tet cells for GFP or Prox1 (green), CyclinE1 (red) and counterstained for DAPI. (h) Quantification of the data (GFP: 58.00±1.51%, Prox1: 78.44±3.86%; Po0.01). The difference in the fold of CyclinE1 induction between western blot and immunostainings could be explained by the strong effect of Prox1 in enhancing the levels of CyclinE1 per individual cell, which is also evident in Figure 7g (note the correlation between high levels of Prox1 induction and strong staining for CyclinE1). (i) Relative expression levels of CyclinE1 mRNA in GFP-Tet and Prox1- Tet cells in the presence of Dox, measured with quantitative real-time RT–PCR. (j) Transcriptional assays in GFP-Tet and Prox1-Tet cells in the presence of Dox, transfected with a luciferase reporter construct containing human CyclinE1 promoter. (k) Immunostaining in the GFP-Tet and Prox1-Tet tumors with GFP or Prox1 (green), CyclinD1 (red) and counterstained for DAPI. (l) Quantification of the data (GFP: 65.23±3.06%, Prox1: 21.28±2.53%; Po0.01). (m) Immunostainings in the GFP-Tet and Prox1-Tet tumors with GFP or Prox1 (green), CyclinE1 (red) and counterstained for DAPI. (n) Quantification of the data (GFP: 59.33±2.47%, Prox1: 79.20±3.28%; Po0.01). Scale bars for (e), (g) and (k): 50 mm. Scale bar for m:25mm.

& 2013 Macmillan Publishers Limited Oncogene (2013) 947 – 960 Prox1 blocks cell cycle progression in neuroblastoma cells IP Foskolou et al 956

Figure 8. Prox1 affects the expression of key cell cycle regulators. (a) Western blot analysis of protein extracts from GFP-Tet and Prox1-Tet cells in the presence of Dox for p27-Kip1, p21-Cip1, p53, Cdc25A and Actin, as indicated, in 48 and 72 h. (b, c) Quantiﬁcation of the p27-Kip1 and Cdc25A blots, respectively. For all cases: Po0.01. (d) Immunostaining in GFP-Tet and Prox1-Tet cells for GFP or Prox1 (green) and p27-Kip1 (red). Scale bar: 50 mm. (e, f) Relative expression levels of p27-Kip1 and Cdc25A mRNAs in GFP-Tet and Prox1-Tet cells in the presence of Dox, measured with quantitative real-time RT–PCR. For p27-Kip1: Po0.01; For Cdc25A: Po0.05. (g) Transcriptional assays in GFP-Tet and Prox1-Tet cells in the presence of Dox, transfected with a luciferase reporter construct containing human Cdc25A promoter Po0.05.

raised the possibility that Prox1 may act as a tumor suppressor in vivo. Prox1 blocks neuroblastoma cell cycle in G0/G1 phase, gene in neuroblastoma cancer. Accordingly, in this study, we show being able to downregulate CyclinD1, CyclinA and CyclinB1. In that Prox1 strongly inhibits the ability of neuroblastoma cells to agreement, it has been previously reported that Prox1 is sufﬁcient proliferate, whereas endogenous expression levels are necessary to suppress proliferation in various cell lines or primary cells from for the proper regulation of cell proliferation both in vitro and other tissues, including hematopoetic stem/progenitor cells,27,28

Oncogene (2013) 947 – 960 & 2013 Macmillan Publishers Limited Prox1 blocks cell cycle progression in neuroblastoma cells IP Foskolou et al 957 neural precursor cells,5,29,30 HEK293, pancreatic cell lines (that is, miapaca2),16 Hep3B and Huh7cells,12 and KYSE esophageal cancer cell line.31 Consistently, decreased expression levels, silencing and/ or loss-of-function mutations for Prox1 have been found in various cancers, including hematological malignacies, hepatocellular, bilary duct, pancreatic and breast cancers.11–16,32 However, the detailed molecular mechanism of Prox1-mediated antiproliferative action was not known. Here, we provide functional evidence that Prox1 acts through a combined action in inducing the cdk inhibitor protein p27-Kip1 and suppressing Cdc25A phosphatase to block cell proliferation. Most important, by alleviating these effects of Prox1 on p27-Kip1 and Cdc25A, we were able to rescue the Prox1 antiproliferative function, indicating that these actions are of functional importance and mediate the Prox1 effect on cell cycle regulation. We propose that Prox1 is directly involved in the mechanisms that control the progression of cell cycle from G1 to S phase (Figure 10m). In particular, we suggest that Prox1, through the regulation of p27-Kip1 and Cdc25A expression, blocks the activity of Cdk2/CyclinE complex, being thus able to inhibit the progression of cell cycle (Figure 10m). However, based on our data we cannot exclude additional actions of these cell cycle regulators in relation to Prox1 effect. In addition, Prox1 appears to control the expression of these regulators at the transcriptional level, possibly through direct binding at their promoters. However, based on our observations we cannot totally rule out the possibility of an indirect mode of regulation and thus the detailed molecular mechanism is currently under investigation. In striking accordance, Drosophila neuroblasts lacking Prospero, the Drosophila homolog of Prox1, form tumors in both the embryonic nervous system and larval brain,21,33–35 being thus considered a tumor suppressor gene. Prospero suppresses the genetic program for self-renewal and cell cycle progression of Drosophila neuroblasts by affecting the expression of String (Cdc25A in mammals), Dacapo (p27-Kip1 in mammals), CycA, CycE, Encore, Rbf (retinoblastoma-family protein in mammals) and E2F genes.21,22,36–38 Interestingly, similar to our data it was recently shown that Prospero inhibits cell cycle progression of neuronal progenitors in Drosophila by activating the expression of Dacapo.36 These observations suggest a conserved mechanism for Prox1-mediated suppression of cell cycle progression from flies to mammals. Although these studies indicate a tumor suppressor function of Prox1, recent reports suggest its involvement in tumor progression and enhancement of proliferation in other cell types and tissues. Thus, it appears that Prox1 positively regulates the cell Figure 9. Prox1 directly interacts with the promoters of Cdkn1b (p27- Kip1) and Cdc25A genes. (a) Schematic representation of the 50 of cycle in lymphatic endothelial cells, fetal hepatic stem/progenitor cells and promotes dysplasia in colonic adenomas and colo- Cdkn1b gene locus. The first two exons of Cdkn1b (p27-Kip1) gene 17,20,39,40 are represented as black boxes. (b) Chromatin immunoprecipitation rectal cancer progression. These data underscore the (ChIP) analysis of the binding sites of Prox1 to Cdkn1b gene locus. complexity of Prox1 actions in cancer pathogenesis and cell cycle ChIP experiments were performed using anti-Prox1 antibody in regulation. In all these cases, Prox1 is able to induce the chromatin isolated from Prox1-Tet cells cultured in the presence of expression of CyclinE1, and especially for lymphatic endothelial Dox, and compared with GFP-Tet cells, as indicated. For Prox1 and cells it was shown that Prox1 exerts a direct effect on activating GFP reactions the same amount of DNA was used as template. The the CyclinE1 promoter.20 Consistently, despite the strong primer pairs used to amplify the corresponding DNA sequences are antiproliferative effect of Prox1 in neuroblastoma cells, it is also indicated with arrows below the schematic drawing in (a). Note that Prox1 specifically binds to the sites B and C of the Cdkn1b locus. sufficient to induce CyclinE1 mRNA expression and promoter (c) Schematic representation of the 50 of Cdc25A gene locus. The first activation similar to lymphatic endothelial cells. However, in N2A, two exons of Cdc25A gene are represented as black boxes. (d) ChIP Prox1 exerts additional actions by activating p27-Kip1 and analysis of the binding sites of Prox1 to Cdc25A gene locus. ChIP suppressing Cdc25A expression that manage to counteract experiments were performed using anti-Prox1 antibody in chroma- CyclinE’s effect on proliferation and further block cell cycle tin isolated from Prox1-Tet cells cultured in the presence of Dox, and progression. This scenario is supported by data suggesting that compared with GFP-Tet cells, as indicated. For Prox1 and GFP p27-Kip1 and Cdc25A regulate cell cycle through their ability to reactions the same amount of DNA was used as template. The inactivate or activate, respectively, the Cdk2/CyclinE complex that primer pairs used to amplify the corresponding DNA sequences are in its active form promotes progression of the cell cycle from G1 to indicated with arrows below the schematic drawing in (c). Note that Prox1 specifically binds to the sites C and D of Cdc25A gene locus. S phase (Figure 10m). Moreover, we suggest that the ability of Prox1 to induce CyclinE1 may account for the context-dependent oncogenic function of Prox1. According to our model, in these cells the Prox1-mediated induction of CyclinE1 can override its

& 2013 Macmillan Publishers Limited Oncogene (2013) 947 – 960 Prox1 blocks cell cycle progression in neuroblastoma cells IP Foskolou et al 958

effect on p27-Kip1 and Cdc25A, or alternatively Prox1 may not be Interferon-g in esophageal cancer cells.31 Interestingly, Interferon- sufﬁcient to affect their expression and hence the CyclinE’s effect g reduces proliferation and delay tumorigenicity in human on proliferation prevails. Again in striking agreement, Prospero in neuroblastoma cells.42,43 Therefore, Prox1 is a candidate factor glial precursors of Drosophila is required to maintain the mitotic that could mediate this effect in neuroblastoma cells. Moreover, potential of glia by positively regulating CyclinE expression and Prox1 expression is negatively regulated by miRNA-181 in human antagonizing the p27-Kip1 homolog, Dacapo,41 in contrast to endothelial cells by direct binding to its 30-untranslated mRNA.44 embryonic neuroblasts, where it suppresses CyclinE expression Strikingly, miRNA-181 appeared to be overexpressed and strongly and induces Dacapo. Thus, in Drosophila similar to mammalian associated with unfavorable human neuroblastoma, considered to cells, Prospero affects cell cycle genes and can either promote be an oncogenic miRNA.45,46 The ability of miRNA-181 to suppress or inhibit them depending on the cellular or developmental Prox1 may also mediate its oncogenic function in neuroblastoma. context. These observations indicate that cytokines, drugs and/or miRNA-/ A key question arising from our observations is whether siRNA-based approaches could be utilized to de-repress Prox1 the ability of Prox1 to block neuroblatoma cell proliferation expression in neuroblastoma cells to promote cell cycle arrest and can be utilized in therapeutic strategies. Accordingly, it was contribute to therapeutic approaches. Towards these aims, we are recently shown that Prox1 mediates the antiproliferative action of currently focussing on identifying drugs and/or pathways that

Oncogene (2013) 947 – 960 & 2013 Macmillan Publishers Limited Prox1 blocks cell cycle progression in neuroblastoma cells IP Foskolou et al 959

Figure 10. Overexpression of Cdc25A and shRNA for p27-Kip1 rescue the antiproliferative effect of Prox1. (a) N2A cells co-transfected with GFP or Prox1 and Cdc25A in a ratio of (1:3), were pulsed for 1 h with BrdU and then labeled for GFP or Prox1 (green), BrdU (red) and 4,6-diamidino- 2-phenylindole (DAPI) (blue). In all cases arrows indicate double-positive cells. (b) Quantification of the data (GFP þ Cdc25A: 23.27±0.97%, Prox1: 1.78±0.75%, Prox1 þ Cdc25A: 7.48±0.66%; Po0.01 for Prox1 vs Prox1 þ Cdc25A). (c) N2A cells co-transfected with GFP or Prox1 and Cdc25A in a ratio of (1:3), were labeled for GFP or Prox1 (green), pH3 (red) and DAPI (blue). (d) Quantification of the data (GFP þ Cdc25A: 10.84±0.44%, Prox1: 1.04±0.19%, Prox1 þ Cdc25A: 3.61±0.66%; Po0.05 for Prox1 vs Prox1 þ Cdc25A). (e) N2A cells co-transfected with GFP or Prox1 and shp27-Kip1 or shLUC in a ratio of (1:3) were pulsed for 1 h with BrdU and then labeled for GFP or Prox1 (green), BrdU (red) and DAPI (blue). (f) Quantification of the data (GFP þ shp27-Kip1: 29.96±0.89%, Prox1 þ shLUC: 1.45±0.1%, Prox1 þ shp27-Kip1: 14.49±0.65%; Po0.01 for Prox1 þ shLUC vs Prox1 þ shp27-Kip1). (g) N2A cells co-transfected with GFP or Prox1 and shp27 or shLUC in a ratio of (1:3), were labeled for GFP or Prox1 (green), pH3 (red) and DAPI (blue). (h) Quantification of the data (GFP þ shp27-Kip1: 7.78±1.05%, Prox1 þ shLUC: 1.17±0.12%, Prox1 þ shp27-Kip1: 4.58±0.47%;Po0.01for Prox1 þ shLUC vs Prox1 þ shp27-Kip1). (i) N2A cells co-transfected with GFP or Prox1, Cdc25A and shp27-Kip1 in a ratio of (1:2.5:2.5), were pulsed for 1 h with BrdU and then labeled for GFP or Prox1 (green), BrdU (red) and DAPI (blue). (j) Quantification of the data (GFP þ Cdc25A þ shp27-Kip1: 22.09±1.56%, Prox1: 1.78±0.75%, Prox1 þ Cdc25A þ shp27-Kip1: 19.55±0.48%; P40.1 for GFP þ Cdc25A þ shp27-Kip1 vs Prox1 þ Cdc25A þ shp27-Kip1). (k) N2A cells co-transfected with GFP or Prox1, Cdc25A and shp27 in a ratio of (1:2.5:2.5), were labeled for GFP or Prox1 (green), pH3 (red) and DAPI (blue). (l) Quantification of the data (GFP þ Cdc25A þ shp27-Kip1: 19.17±0.57%, Prox1: 0.98±0.21%, Prox1 þ Cdc25A þ shp27-Kip1: 15.39±0.48%). For all panels scale bars: 50 mm. (m) Schematic representation of the role of Prox1 in suppressing the cell cycle progression in neuroblastoma cells. Initially, Cyclin/Cdk complexes are kept in an inactive state by phosphorylation. Phosphorylation of Cdk2 is used to accumulate and maintain Cdk2/CyclinE complexes in a pre-active stage.51 Conversely, Cdc25A dual-specificity phosphatase, dephosphorylates and thereby activates Cdk2 and thus induces progression in the cell cycle from G1 to S phase.23,52,53 Here, we demonstrate that Prox1 downregulates expression of the Cdc25A on the mRNA and protein level and thus may impair transition from G1 to S phase. Moreover, Prox1 is also sufficient to induce p27-Kip1 expression on both mRNA and protein level. The cyclin-dependent kinase inhibitor p27-Kip1 is a key regulator of the cell cycle in mammalian cells. Many signal transduction networks regulate p27-Kip1, and it integrates diverse signals into a final decision between proliferation and cell-cycle exit.54 p27-Kip1 is an atypical tumor suppressor that negatively regulates G1–S cell cycle progression by directly inhibiting Cdk2/CyclinE complex.55–57 In G0 and early G1, p27-Kip1 translation and protein stability are maximal and it binds and inhibits Cdk2/CyclinE.58,59 The progressive decrease in p27-Kip1 during G1 allows Cdk2/CyclinE and Cdk2/CyclinA to activate the transcription of genes that are required for the G1–S transition and to participate in the initiation of DNA replication.60Moreover, Prox1 induces CyclinE1 expression, indicating a dual mode of action that could explain the context-dependent oncogenic function of Prox1 in other tissues. We suggest that the balance between the effect of Prox1 on CyclinE and that on p27-kip1/Cdc25A regulates its function in tumor progression or suppression, respectively. can enhance Prox1 expression in neuroblastomas using high- CONFLICT OF INTEREST throughput approaches. The authors declare no conflict of interest.

MATERIALS AND METHODS ACKNOWLEDGEMENTS Cell culture, allotransplantations, transfections, viral infections, luciferase assays, chromatin immunoprecipitation assays, We are grateful to Yukiko Gotoh, Argyris Efstratiadis, Kurt Engeland, Ingrid Hoffman, Fluorescence-activated cell sorting analyses and reagents Aristidis Charonis, Valeria Kaltezioti, Chrisiida Tsimplouli, Kostas Vekrellis, Katerina Melachroinou, Elena Katsantoni, Anthony Gavalas, Dimitra Mangoura, Maria Neuroblastoma cell lines were cultured and transfected as previously 5,47–49 Roubelakis, Eleni Siapati, Stamatis Pagakis, BIU facility of BRAFAA and Addgene for described. Allotransplantations, plasmids, viral infections, luciferase plasmids, reagents, discussions and technical help. assays, chromatin immunoprecipitation assays, fluorescence-activated cell sorting (FACS) analysis, reagents and stable cell lines construction are described in Supplementary Materials and Methods. All animals were handled in strict accordance with good animal practice as defined by the REFERENCES relevant European and Greek animal welfare bodies. 1 Maris JM, Hogarty MD, Bagatell R, Cohn SL. Neuroblastoma. Lancet 2007; 369: 2106–2120. Immunohistochemistry and western blot analysis 2 Pearson AD, Pinkerton CR, Lewis IJ, Imeson J, Ellershaw C, Machin D. High- Prox1 was detected using a rabbit polyclonal (ReliaTech, Wolfenbo¨ttel, dose rapid and standard induction chemotherapy for patients aged over Germany), mouse monoclonal (Millipore, Bedford, MA, USA) or an anti- 1 year with stage 4 neuroblastoma: a randomised trial. Lancet Oncol 2008; 9: FLAG antibody from Sigma (St Louis, MO, USA).5 Anti-BrdU was detected 247–256. using a mouse monoclonal (Dako, Glostrup, Denmark) or a rat antibody 3 Becker J, Wang B, Pavlakovic H, Buttler K, Wilting J. Homeobox transcription factor (Abcam, Cambridge, UK). Anti-GFP from Sigma; PCNA from DAKO; anti-pH3 Prox1 in sympathetic ganglia of vertebrate embryos: correlation with human from Abcam; anti-p27-Kip1 from BD Biosciences (San Jose, CA, USA); anti- stage 4s neuroblastoma. Pediatr Res 2010; 68: 112–117. Notch1, -p53, -p21-Cip1, -CyclinA, -CyclinD1, -CyclinE1 and -Cdc25A from 4 Wigle JT, Harvey N, Detmar M, Lagutina I, Grosveld G, Gunn MD et al. An essential Santa Cruz (Santa Cruz, CA, USA); all antibodies were detected as role for Prox1 in the induction of the lymphatic endothelial cell phenotype. EMBO previously described.5 Immunofluorescent specimens were viewed and J 2002; 21: 1505–1513. analyzed with a Leica (Wetzlar, Germany) confocal microscope. Statistical 5 Kaltezioti V, Kouroupi G, Oikonomaki M, Mantouvalou E, Stergiopoulos A, Char- analysis was performed with the two-tailed paired Student’s t-test. onis A et al. Prox1 regulates the notch1-mediated inhibition of neurogenesis. PLoS Quantification of the digital images obtained was performed using Biol 2010; 8: e1000565. ImageJ software (NIH, Bethesda, MD, USA). 6 Chang HH, Lee H, Hu MK, Tsao PN, Juan HF, Huang MC et al. Notch1 expression predicts an unfavorable prognosis and serves as a therapeutic target of patients with neuroblastoma. Clin Cancer Res 2010; 16: 4411–4420. Proliferation, growth and apoptosis assays 7 Risebro CA, Searles RG, Melville AA, Ehler E, Jina N, Shah S et al. Prox1 maintains Proliferation and apoptotis assays were performed as previously muscle structure and growth in the developing heart. Development 2009; 136: described 5,47–50 and Supplementary Materials and Methods. 495–505. 8 Sosa-Pineda B, Wigle JT, Oliver G. Hepatocyte migration during liver development requires Prox1. Nat Genet 2000; 25: 254–255. RNA extraction and real-time RT–PCR analysis 9 Wigle JT, Chowdhury K, Gruss P, Oliver G. Prox1 function is crucial for mouse lens- Total RNA was isolated by TRIZOL reagent (Ambion) followed by treatment fibre elongation. Nat Genet 1999; 21: 318–322. with RQ1 DNase (Promega). Quantitative RT–PCR analysis was performed as 10 Wigle JT, Oliver G. Prox1 function is required for the development of the murine described.5 Primer sets are listed in Supplementary Materials and Methods. lymphatic system. Cell 1999; 98: 769–778.

& 2013 Macmillan Publishers Limited Oncogene (2013) 947 – 960 Prox1 blocks cell cycle progression in neuroblastoma cells IP Foskolou et al 960 11 Schneider M, Buchler P, Giese N, Giese T, Wilting J, Buchler MW et al. Role of 35 Lee CY, Wilkinson BD, Siegrist SE, Wharton RP, Doe CQ. Brat is a Miranda cargo lymphangiogenesis and lymphangiogenic factors during pancreatic cancer pro- protein that promotes neuronal differentiation and inhibits neuroblast self- gression and lymphatic spread. Int J Oncol 2006; 28: 883–890. renewal. Dev Cell 2006; 10: 441–449. 12 Shimoda M, Takahashi M, Yoshimoto T, Kono T, Ikai I, Kubo H. A homeobox 36 Colonques J, Ceron J, Reichert H, Tejedor FJ. A transient expression of Prospero protein, prox1, is involved in the differentiation, proliferation, and prognosis in promotes cell cycle exit of Drosophila postembryonic neurons through the reg- hepatocellular carcinoma. Clin Cancer Res 2006; 12(20 Pt 1): 6005–6011. ulation of Dacapo. PLoS One 2011; 6: e19342. 13 Nagai H, Li Y, Hatano S, Toshihito O, Yuge M, Ito E et al. Mutations and aberrant 37 Liu TH, Li L, Vaessin H. Transcription of the Drosophila CKI gene dacapo is DNA methylation of the PROX1 gene in hematologic malignancies. Genes Chro- regulated by a modular array of cis-regulatory sequences. Mech Dev 2002; 112: mosomes Cancer 2003; 38: 13–21. 25–36. 14 Versmold B, Felsberg J, Mikeska T, Ehrentraut D, Kohler J, Hampl JA et al. Epi- 38 Southall TD, Brand AH. Neural stem cell transcriptional networks highlight genes genetic silencing of the candidate tumor suppressor gene PROX1 in sporadic essential for nervous system development. EMBO J 2009; 28: 3799–3807. breast cancer. Int J Cancer 2007; 121: 547–554. 39 Kamiya A, Kakinuma S, Onodera M, Miyajima A, Nakauchi H. Prospero-related 15 Dudas J, Mansuroglu T, Moriconi F, Haller F, Wilting J, Lorf T et al. Altered reg- homeobox 1 and liver receptor homolog 1 coordinately regulate long-term ulation of Prox1-gene-expression in liver tumors. BMC Cancer 2008; 8:92. proliferation of murine fetal hepatoblasts. Hepatology 2008; 48: 252–264. 16 Takahashi M, Yoshimoto T, Shimoda M, Kono T, Koizumi M, Yazumi S et al. Loss of 40 Petrova TV, Makinen T, Makela TP, Saarela J, Virtanen I, Ferrell RE et al. Lymphatic function of the candidate tumor suppressor prox1 by RNA mutation in human endothelial reprogramming of vascular endothelial cells by the Prox-1 homeobox cancer cells. Neoplasia 2006; 8: 1003–1010. transcription factor. EMBO J 2002; 21: 4593–4599. 17 Petrova TV, Nykanen A, Norrmen C, Ivanov KI, Andersson LC, Haglund C et al. 41 Grifﬁths RL, Hidalgo A. Prospero maintains the mitotic potential of glial precursors Transcription factor PROX1 induces colon cancer progression by promoting the enabling them to respond to neurons. EMBO J 2004; 23: 2440–2450. transition from benign to highly dysplastic phenotype. Cancer Cell 2008; 13: 407–419. 42 Airoldi I, Meazza R, Croce M, Di Carlo E, Piazza T, Cocco C et al. Low-dose inter- 18 Louvi A, Artavanis-Tsakonas S. Notch signalling in vertebrate neural development. feron-gamma-producing human neuroblastoma cells show reduced proliferation Nat Rev Neurosci 2006; 7: 93–102. and delayed tumorigenicity. Br J Cancer 2004; 90: 2210–2218. 19 Tsarovina K, Schellenberger J, Schneider C, Rohrer H. Progenitor cell maintenance 43 Ribatti D, Nico B, Pezzolo A, Vacca A, Meazza R, Cinti R et al. Angiogenesis in a and neurogenesis in sympathetic ganglia involves Notch signaling. Mol Cell human neuroblastoma xenograft model: mechanisms and inhibition by tumour- Neurosci 2008; 37: 20–31. derived interferon-gamma. Br J Cancer 2006; 94: 1845–1852. 20 Baxter SA, Cheung DY, Bocangel P, Kim HK, Herbert K, Douville JM et al. Reg- 44 Kazenwadel J, Michael MZ, Harvey NL. Prox1 expression is negatively regulated by ulation of the lymphatic endothelial cell cycle by the PROX1 homeodomain miR-181 in endothelial cells. Blood 2010; 116: 2395–2401. protein. Biochem Biophys Acta 2011; 1813: 201–212. 45 Chen Y, Stallings RL. Differential patterns of microRNA expression in neuro- 21 Choksi SP, Southall TD, Bossing T, Edoff K, de Wit E, Fischer BE et al. Prospero acts blastoma are correlated with prognosis, differentiation, and apoptosis. Cancer Res as a binary switch between self-renewal and differentiation in Drosophila neural 2007; 67: 976–983. stem cells. Dev Cell 2006; 11: 775–789. 46 Schulte JH, Marschall T, Martin M, Rosenstiel P, Mestdagh P, Schlierf S et al. Deep 22 Li L, Vaessin H. Pan-neural Prospero terminates cell proliferation during Droso- sequencing reveals differential expression of microRNAs in favorable versus phila neurogenesis. Genes Dev 2000; 14: 147–151. unfavorable neuroblastoma. Nucleic Acids Res 2010; 38: 5919–5928. 23 Rother K, Kirschner R, Sanger K, Bohlig L, Mossner J, Engeland K. p53 down- 47 Georgopoulou N, Hurel C, Politis PK, Gaitanou M, Matsas R, Thomaidou D. BM88 is regulates expression of the G1/S cell cycle phosphatase Cdc25A. Oncogene 2007; a dual function molecule inducing cell cycle exit and neuronal differentiation of 26: 1949–1953. neuroblastoma cells via cyclin D1 down-regulation and retinoblastoma protein 24 Timofeev O, Cizmecioglu O, Settele F, Kempf T, Hoffmann I. Cdc25 phosphatases hypophosphorylation. J Biol Chem 2006; 281: 33606–33620. are required for timely assembly of CDK1-cyclin B at the G2/M transition. J Biol 48 Politis PK, Akrivou S, Hurel C, Papadodima O, Matsas R. BM88/Cend1 is involved in Chem 2010; 285: 16978–16990. histone deacetylase inhibition-mediated growth arrest and differentiation of 25 Itoh Y, Masuyama N, Nakayama K, Nakayama KI, Gotoh Y. The cyclin-dependent neuroblastoma cells. FEBS Lett 2008; 582: 741–748. kinase inhibitors p57 and p27 regulate neuronal migration in the developing 49 Politis PK, Makri G, Thomaidou D, Geissen M, Rohrer H, Matsas R. BM88/CEND1 mouse neocortex. J Biol Chem 2007; 282: 390–396. coordinates cell cycle exit and differentiation of neuronal precursors. Proc Natl 26 Rodriguez-Niedenfuhr M, Papoutsi M, Christ B, Nicolaides KH, von Kaisenberg CS, Acad Sci USA 2007; 104: 17861–17866. Tomarev SI et al. Prox1 is a marker of ectodermal placodes, endodermal com- 50 Stellas D, Karameris A, Patsavoudi E. Monoclonal antibody 4C5 immunostains partments, lymphatic endothelium and lymphangioblasts. Anat Embryol 2001; human melanomas and inhibits melanoma cell invasion and metastasis. Clin 204: 399–406. Cancer Res 2007; 13: 1831–1838. 27 Hope KJ, Cellot S, Ting SB, MacRae T, Mayotte N, Iscove NN et al. An RNAi screen 51 Hassepass I, Voit R, Hoffmann I. Phosphorylation at serine 75 is required for UV- identiﬁes Msi2 and Prox1 as having opposite roles in the regulation of hemato- mediated degradation of human Cdc25A phosphatase at the S-phase checkpoint. poietic stem cell activity. Cell Stem Cell 2010; 7: 101–113. J Biol Chem 2003; 278: 29824–29829. 28 Hope KJ, Sauvageau G. Roles for MSI2 and PROX1 in hematopoietic stem cell 52 Nilsson I, Hoffmann I. Cell cycle regulation by the Cdc25 phosphatase family. Prog activity. Curr Opin Hematol 2011; 18: 203–207. Cell Cycle Res 2000; 4: 107–114. 29 Dyer MA, Livesey FJ, Cepko CL, Oliver G. Prox1 function controls progenitor cell 53 Sexl V, Diehl JA, Sherr CJ, Ashmun R, Beach D, Roussel MF. A rate limiting function proliferation and horizontal cell genesis in the mammalian retina. Nat Genet 2003; of cdc25A for S phase entry inversely correlates with tyrosine dephosphorylation 34: 53–58. of Cdk2. Oncogene 1999; 18: 573–582. 30 Elkouris M, Balaskas N, Poulou M, Politis PK, Panayiotou E, Malas S et al. Sox1 54 Chu IM, Hengst L, Slingerland JM. The Cdk inhibitor p27 in human cancer: maintains the undifferentiated state of cortical neural progenitor cells via the prognostic potential and relevance to anticancer therapy. Nat Rev Cancer 2008; 8: suppression of Prox1-mediated cell cycle exit and neurogenesis. Stem Cells 2011; 253–267. 29: 89–98. 55 Grana X, Reddy EP. Cell cycle control in mammalian cells: role of cyclins, cyclin 31 Akagami M, Kawada K, Kubo H, Kawada M, Takahashi M, Kaganoi J et al. Tran- dependent kinases (CDKs), growth suppressor genes and cyclin-dependent scriptional Factor Prox1 Plays an Essential Role in the Antiproliferative Action of kinase inhibitors (CKIs). Oncogene 1995; 11: 211–219. Interferon-gamma in Esophageal Cancer Cells. Ann Surg Oncol 2011; 18: 3868–3877. 56 Liu S, Yamauchi H. p27-Associated G1 arrest induced by hinokitiol in human 32 Laerm A, Helmbold P, Goldberg M, Dammann R, Holzhausen HJ, Ballhausen WG. malignant melanoma cells is mediated via down-regulation of pRb, Skp2 ubi- Prospero-related homeobox 1 (PROX1) is frequently inactivated by genomic quitin ligase, and impairment of Cdk2 function. Cancer Lett 2009; 286: 240–249. deletions and epigenetic silencing in carcinomas of the bilary system. J Hepatol 57 Molinari M. Cell cycle checkpoints and their inactivation in human cancer. Cell 2007; 46: 89–97. Prolif 2000; 33: 261–274. 33 Bello B, Reichert H, Hirth F. The brain tumor gene negatively regulates neural 58 Hengst L, Reed SI. Inhibitors of the Cip/Kip family. Curr Top Microbiol Immunol progenitor cell proliferation in the larval central brain of Drosophila. Development 1998; 227:25–41. 2006; 133: 2639–2648. 59 Sherr CJ, Roberts JM. CDK inhibitors: positive and negative regulators of G1-phase 34 Betschinger J, Mechtler K, Knoblich JA. Asymmetric segregation of the tumor progression. Genes Dev 1999; 13: 1501–1512. suppressor brat regulates self-renewal in Drosophila neural stem cells. Cell 2006; 60 Nigg EA. Targets of cyclin-dependent protein kinases. Curr Opin Cell Biol 1993; 5: 124: 1241–1253. 187–193.

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

Oncogene (2013) 947 – 960 & 2013 Macmillan Publishers Limited