Expression Profiling of Ral-Depleted Bladder Cancer Cells Identifies RREB-1 As a Novel Transcriptional Ral Effector

Oncogene (2007) 26, 7143–7152 & 2007 Nature Publishing Group All rights reserved 0950-9232/07 $30.00 www.nature.com/onc ORIGINAL ARTICLE Expression proﬁling of Ral-depleted bladder cancer cells identiﬁes RREB-1 as a novel transcriptional Ral effector

G Oxford1, SC Smith1, G Hampton2 and D Theodorescu1

1Department of Molecular Physiology and Biological Physics, Health Sciences Center, University of Virginia, Charlottesville, VA, USA and 2Genomics Institute of the Novartis Research Foundation, San Diego, CA, USA

Although the monomeric GTPases RalA and RalB have Introduction been shown to regulate a variety of transcription factors, little is known regarding the differences or similarities Within the Ras family, RalA and RalB are closely in transcriptional programs regulated by RalA compared related proteins with both overlapping and distinct to RalB. Further, the association of these transcriptional functions (Feig, 2003). Initially discovered as Ras pathways to human carcinogenesis and progression homologs, they were shown to be regulated, at least in remains unclear. Here, we studied the role of RalA and/ part, by Ras through its ability to stimulate RalGEFs or RalB in transcriptional regulation by combining short such as RalGDS and Rlf (Wolthuis et al., 1998). interfering RNA depletion of Ral withgene expression Although studies on rodent tumorigenesis suggested profiling via microarray in the human bladder cancer cell that RalGEFs, and by extension RalA and RalB, are line, UMUC-3. A large number of genes were found to be not particularly tumorigenic (White et al., 1995), recent similarly modulated in cells withRalA and RalB results indicate more important roles for RalA and RalB depletion, suggesting that RalA and RalB impinge on in human cancer. For example, RalGEFs have been overlapping transcriptional signaling pathways. However, shown to be the most important effectors downstream smaller sets of genes were modulated by depletion of RalA of Ras in the transformation of human cells (Hamad or RalB, indicating that these closely related proteins et al., 2002) and RalA is a necessary component of Ras- also regulate nonoverlapping transcriptional pathways. driven tumorigenesis (Lim et al., 2005). Conversely, Computational analysis of upstream sequences of genes RalB has been shown to be pro-migratory in human modulated by Ral depletion identified Ras-responsive bladder and prostate cancer cells (Oxford et al., 2005), element-binding protein (RREB)-1, as a putative Ral and depletion of RalB via short interfering RNA transcriptional target, which we verified experimentally. (siRNA) has been shown to trigger apoptosis in many Importantly, as a group, Ral-regulated probe sets identi- human cancer cell lines (Chien and White, 2003). Both fied here were disproportionally represented among those RalA and RalB have been shown to be pro-metastatic differentially expressed as a function of human bladder (Ward et al., 2001) and they appear to share an transformation. Taken together, these data strongly overlapping function in regulating cancer cell prolifera- suggest that Ral family members mediate both common tion (Oxford et al., 2005). and specific transcriptional programs that are associated Ral proteins mediate cellular functions via a reper- withhumancancer and identify RREB-1 as a novel toire of effectors such as RalBP1, a GTPase-activating transcriptional effector of Ral. protein for Rac1 and Cdc42 (Jullien-Flores et al., 1995), Oncogene (2007) 26, 7143–7152; doi:10.1038/sj.onc.1210521; which also plays a role in the transport of glutathione published online 14 May 2007 conjugates (Awasthi et al., 2000), Sec5 and exo84, components of the exocyst involved in regulated Keywords: GTPases; bladder neoplasms; biomarker; secretion (Moskalenko et al., 2003), and the actin metastasis regulatory protein filamin (Ohta et al., 1999). Through these, RalA and/or RalB regulate endocytosis of membrane receptors, exocytosis and delivery of polarized membrane proteins (Shipitsin and Feig, 2004), cytoskeletal dynamics (Lebreton et al., 2004), drug resistance (Awasthi et al., 2003), motility (Gildea et al., 2002) and both anchorage-dependent and anchorage-independent proliferation (Chien and White, 2003). Correspondence: Professor D Theodorescu. Department of Molecular In addition, RalA and/or RalB have also been shown Physiology and Biological Physics, Health Sciences Center, University to regulate transcription factors such as nuclear factor k B of Virginia, Box 422, Charlottesville, VA 22908, USA. E-mail: [email protected] (NF-kB), AP-1, FOXO4, HSF-1 and ZONAB (Camonis Received 15 September 2006; revised 25 January 2007; accepted 5 March and White, 2005). However, a fundamentally impor- 2007; published online 14 May 2007 tant question remains in the study of transcriptional RalA and RalB functional expression profiling in human bladder cancer G Oxford et al 7144 regulation by Ral GTPases, namely whether RalA and marked overlap in the set of genes regulated by RalA RalB affect different transcriptional targets and whether and RalB, with many probe sets showing significant this is carried out via the selective use of different expression changes in both RalA- and RalB siRNA- transcription factors. Here, we sought to resolve this transfected cells. Furthermore, there is a larger number issue and to understand how transcriptional regulation of probe sets, which show changes in expression only contributes to RalA and RalB function in human when cells are depleted of both RalA and RalB cancer. These contributions were measured by selective compared to each GTPase alone (547 vs 162), suggesting depletion of RalA and/or RalB by siRNA knockdown that, in many cases, unique signaling pathways may be in human bladder cancer cells and profiling the changes triggered only when both RalA and RalB activity is in gene expression following protein depletion. Com- altered. This indicates that functional interactions mon transcriptional components were identified and between several signaling pathways occur downstream mechanistically validated. of these two proteins. Table 1 shows the top 10 genes with the largest statistically significant positive (upregulated) or negative (downregulated) fold changes in response to siRNA- Results mediated depletion of RalA, RalB or both RalA and RalB. Here also, the combination of both overlapping Expression profiles in response to protein knockdown and unique contributions of RalA and RalB to indicate both similar and unique transcriptional networks transcriptional regulation is evident. Of the top 10 downstream of Ral family members downregulated genes in response to RalA depletion, 6 Because RalA and RalB both play important roles in were also significantly downregulated in response to human cancer and also regulate various transcription RalB depletion. Conversely, 8 of the 10 most down- factors through mostly unknown mechanisms, we chose regulated genes with RalB depletion were also down- to investigate the contributions of RalA and RalB to regulated with RalA depletion. For upregulated genes, 8 overall transcriptional regulation. We chose to focus on out of the top 10 for both RalA and RalB depletion the UMUC-3human bladder cancer cell line because: were also significantly upregulated in the other case. (1) it is derived from an invasive tumor (Grossman et al., Uniquely regulated, in response to RalA depletion, were 1986) and maintains high in vitro migration (Oxford solute carrier family 11, asparagine synthetase (both et al., 2005) and invasion (Wu et al., 2004) activities, (2) upregulated), eukaryotic translation initiation factor 4E it has measurably active RalA and RalB (Oxford et al., binding protein 2 and interleukin 1 receptor accessory 2005), presumably, in part, due to harboring an protein 1 (both downregulated). With RalB depletion, activating K-Ras mutation (Jebar et al., 2005) and (3) both ubiquitin-specific protease 18 and promyelocytic we had previously shown in this cell line that siRNA- leukemia ubiquitin ligase were uniquely upregulated and mediated RalB depletion results in decreased cell natural killer cell transcript 4 was uniquely down- motility and that depletion of both RalA and RalB regulated. impairs cell proliferation (Oxford et al., 2005). We Given that depletion of RalA and/or RalB can trigger performed gene expression analysis using oligonucleo- a wide range of phenotypes in cancer cells, we analysed tide arrays on UMUC-3cells that had been transfected the gene sets regulated in response to RalA and/or RalB with siRNA specific for RalA, RalB or both RalA and siRNA for gene ontology classes. The results of gene RalB. Parallel transfections were processed for either ontology analysis are shown in Table 2, with only one protein or RNA. Figure 1a shows the effect of siRNA ontogeny class identified as statistically significant transfection on Ral protein levels, with both RalA- and (Po0.001) in both the RalA- and RalB-regulated genes, RalB-specific duplexes reducing the expression of their and six somewhat overlapping ontology classes asso- respective targets B90% while having little effect on the ciated with RalA/B double knockdown. expression of the other Ral protein, consistent with previous results (Oxford et al., 2005). Also consistent was the effect of a single siRNA duplex targeting both RalA and RalB signal via different repertoires of RalA and RalB, which inhibited RalB expression transcription factors B90% and RalA expression B65%. Figure 1b indicates Recently, tools have been developed to uncover putative that siRNA-mediated depletion of RalA and/or RalB transcription factor-regulatory pathways associated proteins was accompanied by a parallel reduction in with changes in transcriptional profiles by looking for corresponding messenger RNA levels, as measured on overrepresented transcription factor binding sites in the the oligonucleotide array. regulatory regions of genes with expression changes. We The numbers of differentially expressed probe sets in used such a tool called Computational Ascertainment of response to siRNA are summarized in Figure 1c. These Regulatory Relationships (Inferred from Expression) or data suggest the presence of unique transcriptional CARRIE (Haverty et al., 2004), to analyse the genes pathways regulated by RalA (62 and 15 genes, uniquely regulated in response to Ral depletion. The most down- and upregulated, respectively, in response to significant predictions by CARRIE of transcription RalA siRNA) and RalB (54 and 31 genes, uniquely factors regulated by Ral are shown in Table 3. down- and upregulated, respectively, in response to Strikingly, the most significant prediction by CAR- RalB siRNA). Interestingly, these data also indicate RIE is that NF-kB may be regulated by RalB. It has

Oncogene RalA and RalB functional expression proﬁling in human bladder cancer G Oxford et al 7145 a GL2 RalA I RalB II RalA /B

RalA

RalB

1.4 1.2 RalA 1 RalB 0.8 0.6 0.4 0.2 0 Relative Protein Expression GL2 RalA I RalB II RalA/B siRNA b 1.2

1 RalA 0.8 RalB probe 1

0.6 RalB probe 2

0.4

0.2

0 Relative mRNA Expression RalA I RalB II RalA/B siRNA

c Downregulated Upregulated

Ral A (N = 141) Ral B (N = 140) Ral A (N = 84) Ral B (N = 116)

62 20 54 15 6 31

36 46 23 30 17 33

315 232

Ral AB (N = 404) Ral AB (N = 328)

Figure 1 Protein and messenger RNA evaluation of UMUC-3human bladder carcinoma cells transfected with Ral siRNA. (a) siRNA transfection selectively depletes RalA, RalB or both RalA and RalB at the protein level. Top panel, representative western blot of UMUC-3cells 72 h after siRNA transfection. Bottom panel, quantitative Ral protein levels in siRNA-transfected cells. Results shown are the averages from two separate experiments with Ral protein expression normalized to control (GL2) siRNA transfection in each experiment. Error bars indicate standard deviation. (b) siRNA transfection selectively depletes RalA, RalB or both RalA and RalB at the RNA level. siRNA transfections of UMUC-3cells carried out in parallel to those shown in ( a) above were analysed for gene expression using Affymetrix HG-U133A oligonucleotide arrays. Results shown are the averages from two separate experiments with Ral mRNA amounts normalized to control (GL2) siRNA transfection in each experiment. Note that HG-U133A chips contain one probe set for RalA (214435_x_at) and two probe sets for RalB (202100_at, 202101_s_at). Error bars indicate standard deviation. (c) Venn diagrams indicating numbers of Affymetrix U133A probe sets signiﬁcantly regulated (fold change>2.0 or oÀ2.0, Po0.1) in response to Ral siRNA transfection. SiRNA, short interfering RNA.

previously been shown that both RalA and RalB test the prediction that RalB regulates NF-kBin regulate cyclin D1 expression in NIH 3T3 cells through UMUC-3bladder cancer cells, we generated UMUC-3 activation of NF-kB (Henry et al., 2000). This regula- cells stably expressing a luciferase reporter gene under tion occurs through unknown effectors, but does not the control of an NF-kB-responsive promoter. Figure 2a appear to involve the Ral effectors RalBP1 or PLD1. To shows that the luciferase activity of these cells was

Oncogene RalA and RalB functional expression proﬁling in human bladder cancer G Oxford et al 7146 Table 1 Affymetrix HG-U133A probe sets and gene names differen- Table 1 (continued ) tially expressed as a function of Ral siRNA transfection in human a b bladder cancer cell line UMUC3 Unigene ID Gene description Fold Unigene ID Gene descriptiona Foldb Hs.77326 Insulin-like growth factor binding protein 3 À9.8 Hs.119129 Collagen, type IV, a1 (COL4A1) À7 Downregulated with RalA siRNA transfection (vs GL2) Hs.14368 SH3 domain binding glutamic acid-rich À6.8 Hs.6385 DENN/MADD domain containing 2A À25.2 protein like (SH3BGRL) Hs.278712 Eukaryotic translation initiation factor 4E À13.9 Hs.146428 Collagen, type V, a1 À6.5 binding protein 2 Hs.242271 RAB GTPase activating protein 1-like À6.5 Hs.10235 Chromosome 5 open reading frame 4 À9.3 Hs.279581 Host cell factor C1 regulator 1 (XPO1 À6.3 (C5ORF4) dependent) Hs.77326 Insulin-like growth factor binding protein 3 À8.3 Hs.146428 Collagen, type V, a1 À7.9 Upregulated with RalA/B siRNA transfection (vs GL2) Hs.198278 6-phosphofructo-2-kinasefructose-2,6- À7.6 Hs.77711 Ets variant gene 4 (E1A enhancer-binding 11.2 biphosphatase 4 (PFKFB4) protein, E1AF) Hs.6906 V-ral simian leukemia viral oncogene À7.6 Hs.25674 Methyl-CpG binding domain protein 2 7.3 homolog A (RALA) Hs.7764 Kelch-like 21 5.8 Hs.209065 KIAA0319-like protein À6.9 Hs.54413SMYD family member 5 5.3 Hs.7407 Epsin 2 À6.3 Hs.83169 Matrix metalloproteinase 1 (interstitial 5.3 Hs.173880 Interleukin 1 receptor accessory protein À6.1 collagenase) (MMP1) (IL1RAP)1 Hs.108196 DNA replication complex GINS protein 5.2 PSF2 Upregulated with RalA siRNA transfection (vs GL2) Hs.30464 Cyclin E2 splice variant 1 5.1 Hs.284235 Solute carrier family 7, (cationic amino acid 4.9 Hs.91985 Wingless-type MMTV integration site 5 transporter) member 11 family, member 10B (WNT10B) Hs.83169 Matrix metalloproteinase 1 (interstitial 4 Hs.84318 Replication protein A1 (70 kDa) 4.6 collagenase) (MMP1) Hs.315167 Defective in sister chromatid cohesion 4.6 Hs.84318 Replication protein A1 (70 kDa) 3.7 homolog 1 Hs.110695 Splicing factor 3b, subunit 5, 10 kDa 3.6 Hs.108196 DNA replication complex GINS protein 3.2 aAffymetrix (www.Affymetrix.com). bAverage fold expression calcu- PSF2 lated by dividing average probe expression value of Ral siRNA- Hs.84318 Replication protein A1 (70 kDa) (RPA1) 3.2 transfected cells by that of GL2-transfected cells. If o1, negative Hs.75692 Asparagine synthetase (ASNS) 3.2 reciprocal was calculated to indicate decreased value. cProbe sets for Hs.75765 GRO2 oncogene 3.1 RalA and RalB were down regulated (Figure 1b) but were not among Hs.180383 Dual speciﬁcity phosphatase 6 3.1 top 10 list. Hs.87246 BCL2 binding component 33.1

Downregulated with RalB siRNA transfection (vs GL2) Hs.943Natural killer cell transcript 4 (NK4) À10.1 responsive to tumor necrosis factor a, consistent with Hs.6385 DENN/MADD domain containing 2A À9.2 Hs.146428 Collagen, type V, a1 À8.5 this promoter being responsive to NF-kB (Henry et al., Hs.75106 Clusterin (complement lysis inhibitor, À7.9 2000). When these cells were transfected with siRNA apolipoprotein J) targeting RalA and/or RalB, only the cells with RalB Hs.211573Heparan sulfate proteoglycan (HSPG2) À6.3 siRNA showed a significant decrease (B30%) in Hs.146428 Collagen, type V, a1 À4.9 Hs.348024 V-ral simian leukemia viral oncogene À4.7 luciferase activity (Figure 2b). These data underscore homolog B (RalB) the ability of CARRIE to accurately predict transcrip- Hs.6467 Synaptogyrin 3(SYNGR3) À4.6 tion factor pathways regulated in response to siRNA Hs.77899 Tropomyosin 1 (a) (TPM1) À4.3 knockdown. To our knowledge, this is the first valida- Hs.150443Talin 2 À4.2 tion of this software program in mammalian cells. Upregulated with RalB siRNA transfection (vs GL2) CARRIE also predicted that the Ras-responsive Hs.83169 Matrix metalloproteinase 1 (interstitial 7.6 element-binding protein (RREB)-1 is regulated by the collagenase) (MMP1) Ral GTPases. RREB-1 was discovered as a zinc-finger Hs.180383 Dual specificity phosphatase 6 7.1 protein involved in mediating a Ras-driven differentia- Hs.38260 Ubiquitin specific protease 18 (USP18) 5.7 Hs.89633 Promyelocytic leukemia ubiquitin ligase 5.5 tion program in a thyroid carcinoma cell line (Thiaga- Hs.180383 Dual specificity phosphatase 6, 5 lingam et al., 1996). Given that both Ral and RREB-1 Hs.84318 Replication protein A1 (70 kDa) 4.4 are regulated by Ras, we sought to determine the Hs.8575 Dihydrouridine synthase 2-like 4.4 accuracy of the CARRIE prediction. We generated Hs.129943Signal-induced proliferation-associated 1 4 like 3 UMUC-3cells stably expressing an RREB-1-driven Hs.108196 DNA replication complex GINS protein 3.9 luciferase reporter gene and showed that its expression PSF2 was reduced by expression of an activated (G12V) H- Hs.110695 Splicing factor 3b, subunit 5, 10 kDa 3.8 Ras allele (Figure 3a). This is consistent with RREB-1 being a transcriptional repressor of p16/INK4a (Zhang Downregulated with RalA/B siRNA transfection (vs GL2)c Hs.211573Heparan sulfate proteoglycan (HSPG2) À37.4 et al., 2003). Figure 3b shows that transfection of siRNA Hs.75106 Clusterin (complement lysis inhibitor, À18.2 targeting either RalB or RalA/B reduced luciferase apolipoprotein J) expression in these RREB-1 reporter cells by 35–45%. Hs.77326 Insulin-like growth factor binding protein 3 À16.2 In contrast, RalA siRNA appeared to have no effect Hs.118787 Transforming growth factor, beta-induced, À11.8 68 kDa (TGFBI) on the RREB-luciferase reporter. To further test the regulation of RREB-1 by RalA and/or RalB, we

Oncogene RalA and RalB functional expression proﬁling in human bladder cancer G Oxford et al 7147 Table 2 Ontology classiﬁcation of genes differentially expressed as a function of Ral siRNA transfection (Pp0.001) Ontology classa Number of genes in class/number of Total genes P-valueb genes measured

Regulated with RalA siRNA transfection (vs GL2) (n ¼ 225) Response to stress 18/126 1230/28042 0.00016

Regulated with RalB siRNA transfection (vs GL2) (n ¼ 256) Protein binding 46/119 4994/28042 0.000004

Regulated with RalA/B siRNA transfection (vs GL2) (n ¼ 732) Protein binding 146/400 4994/28042 2.4 Â 10À19 Intracellular organelle 196/400 8909/28042 6.0 Â 10À11 Regulation of growth 11/400 141/28042 0.00025 Regulation of cell growth 10/400 130/28042 0.00067 Muscle development 11/400 163/28042 0.00087 Cytoskeletal protein binding 18/400 476/28042 0.001 aGene-Ontology database (GO: http://www.geneontology.org). bCalculated by GOstat (http://gostat.wehi.edu.au).

Table 3 Putative transcription factors regulated by Ral GTPases 607 probe sets that were differentially expressed in P-valuea human bladder cancer vs normal tissue. Of these 607 probe sets, we identified a total of 84 that were also RalA siRNA transfection (vs GL2) regulated in response to RalA and/or RalB knockdown PPAR-gamma/RXR-a 1.29 Â 10À7 (Table 4A). Further analysis showed that the overlaps RalB siRNA transfection (vs GL2) between probe sets differentially regulated in bladder NF-kB 1.02 Â 10À15 cancer and those regulated by RalA siRNA, RalB CF2-II 9.96 Â 10À15 siRNA or RalA/B siRNA were, in all cases, statistically FREAC-7 2.34 Â 10À8 significant (Table 4b in Supplementary Materials). À7 S8 7.35 Â 10 These results indicate that the Ral-regulated genes are RalA/B siRNA transfection (vs GL2) overrepresented in the set of bladder cancer genes, RREB-1 1.06 Â 10À10 highlighting the importance of Ral GTPases in human E2F 1.26 Â 10À7 bladder cancer transformation. aOnly factors with Po10 shown. Calculated by CARRIE Web Service (http://zlab.bu.edu/CarrieServer/html/). Discussion transfected activated (G23V) versions of RalA and/or Recent studies on the roles of RalA and RalB in cancer RalB into the RREB-luciferase reporter cells. Figure 3c have pointed toward an antagonistic relationship. For shows that both activated RalA and RalB increased example, in bladder and prostate cancer cells, RalB RREB-responsive luciferase activity by B40%, in both stimulates migration, whereas RalA is antimigratory continuous serum and serum-starved conditions. These (Oxford et al., 2005). In a defined model of Ras-driven results validated the CARRIE prediction of RREB-1 tumorigenesis of human cells, RalA has been shown to regulation by RalA and RalB, thus revealing a novel Ral promote transformation and tumor formation, while effector. In Figure 4a, the stimulation of RREB-1 RalB has been shown to inhibit both (Lim et al., 2005). activity by RalA and RalB is demonstrated in four other RalB depletion was also shown to trigger apoptosis in human cancer cell lines, including two bladder lines certain cancer cell lines, and this was reversed by (UMUC-6, KU-7), PC-3prostate cancer cells and HeLa simultaneous depletion of RalA and RalB (Chien and cervical cancer cells. Also shown is the inhibition of White, 2003). In light of these examples of antagonism RREB-1 activity by H-Ras in all the four cell lines. between RalA and RalB, it is notable that expression of Having elucidated the sets of genes regulated in very few genes appeared to be regulated in an opposing response to RalA and/or RalB depletion in a human manner by RalA vs RalB depletion (data not shown). bladder cancer cell line, we sought to ground these This suggests that many, if not most, of the functional in vivo by determining the extent to which the RalA/ antagonism between RalA and RalB does not involve RalB-driven transcriptional regulation contributes to transcriptional regulation. human bladder cancer. To do this, we took advantage of In contrast to the functional antagonism between 2 previously published data sets of gene expression RalA and RalB described above, depletion of RalA or profiling of human bladder tumors (Dyrskjot et al., RalB results in similar modulation of cellular transcrip- 2004; Nicholson et al., 2004), which, when combined, tion. In response to specific depletion, expression of result in a cohort of 65 bladder tumors and 20 normal many probe sets is modulated not only in the same bladder epithelial samples profiled on HG-U133A direction, but also with roughly the same magnitude. GeneChips. From this cohort, we identified a total of One way of interpreting the large number of genes that

Oncogene RalA and RalB functional expression proﬁling in human bladder cancer G Oxford et al 7148 a 0.5 serum 0.45 serum starved 0.4 * 0.35

0.3

0.25

0.2 Normalized RLU 0.15

0.1

0.05

0 01050 TNF-alpha (ng /mL) b RalA

RalB

0.2

0.15 * 0.1

0.05 Normalized RLU

0 controlRalA I RalB II RalA /B siRNA Figure 2 RalB Regulates NF-kB in UMUC-3cells. ( a) UMUC-3stably expressing an NF- kB-luciferase reporter responding to TNF. Data shown are the averages of duplicate experiments, normalized to cell number using CyQUANT, with error bars indicating standard deviation. (*, Po0.05, t-test). (b) NF-kB-driven luciferase activity in UMUC-3cells transfected with Ral siRNA. Data shown are the averages of duplicate experiments, normalized to cell number using CyQUANT, with error bars indicating standard deviation. (*, Po0.05, t-test). Inset, western blot indicating RalA and RalB protein levels in siRNA-transfected cells from representative parallel transfection. NF-kB, nuclear factor k B; siRNA, short interfering RNA; TNF, tumor necrosis factor.

are perturbed with the knockdown of both RalA and mediated by RalA. With RalB depletion, the only RalB (Figure 1c) is that much of the ability to regulate significant ontology class was ‘protein binding’. We have transcription is shared. Thus, in the absence of one Ral, previously shown that RalB depletion results in impair- the other is still sufficient to modulate the relevant ment of motility and a reduction in actin stress fibers pathway(s). However, only the depletion of both Ral (Oxford et al., 2005) and others have shown that RalB proteins results in additional changes in gene expression. depletion can trigger apoptosis of some cancer cells The only ontology class associated with RalA (Chien and White, 2003). However, with such a broad depletion was that of ‘genes regulated in response to ontology class as protein binding, it is difficult to stress’ (Table 2). While we have previously shown that ascertain the connections to cellular phenotype. RalA depletion has no effect on motility or proliferation With the codepletion of RalA and RalB, gene in bladder cancer cells, others have shown that both ontology analysis uncovered six significant classes. growth in soft agar and tumorigenic growth of Many of the ontology classes found in genes responding Ras-transformed cells are inhibited by RalA depletion to RalA and RalB depletion, with the exception of (Chien and White, 2003). Growth under such subop- ‘protein binding’, were unique and relevant to the timal conditions may require stress responses that are phenotypes associated with RalA/B knockdown in these

Oncogene RalA and RalB functional expression profiling in human bladder cancer G Oxford et al 7149 cells. For example, we have previously shown that stress fibers in RalA/B-depleted cells (Oxford et al., RalA/B depletion in UMUC-3bladder cancer cells 2005). inhibits proliferation (Oxford et al., 2005), particularly We uncovered two novel transcription factor targets in reduced serum conditions, without triggering apop- under Ral regulation in human bladder cancer cells. tosis. Two of the gene ontology classes uncovered in While RalA and RalB regulation of NF-kB in NIH 3T3 these cells were ‘regulation of growth’ and ‘regulation of cells has previously been noted (Henry et al., 2000), it cell growth’. This suggests that transcriptional regula- appears that RalB is a more potent regulator of NF-kB tion by RalA and RalB plays a significant role in than RalA in UMUC-3cells. This discrepancy may generating this phenotype. The other two gene ontology reflect the differences between Ral/NF-kB activities in classes, ‘muscle development’ and ‘cytoskeletal protein mouse (NIH3T3) vs human cells (UMUC-3), or reflect binding’, may be associated with the total loss of actin different mechanisms in cancer cells. This study is also the first description of RREB-1 regulation by RalA/B. Given that RalGEFs are canonical effectors of Ras, and that Ras is known to a regulate Ral activation through them, it may not be HA-H-RAS surprising that a Ras-responsive transcription factor such as RREB-1 is also regulated by RalA and/or RalB. 0.14 The activation of Ras leading to Ral activation and 0.12 serum RREB-1 regulation could either be positive or negative, serum-starved 0.1 depending on which gene is being studied in a specific cell type. Interestingly, we found that Ral activation 0.08 appears to be opposing Ras activation in the regulation * 0.06 of RREB-1, in that a constitutively active H-Ras allele 0.04 reduced the expression of an RREB-driven reporter

Normalized RLU gene. In contrast, constitutively active RalA and RalB 0.02 increased the expression of the reporter construct. The 0 relative contributions of Ras, RalA/B, as well as other FLAG V12-H-Ras Ras effectors to RREB-1 regulation warrant additional study. RREB-1 has been shown to regulate expression b RalA of the tumor suppressor p16/INK4a (Zhang et al., 2003) and has also been found in a histone-modifying RalB corepressor complex called CtBP (Shi et al., 2003), 0.3 which has been implicated in tumorigenesis. Thus, the relationship between Ral GTPases and RREB-1 in 0.25 human cancer will be an important component of 0.2 further work. * * Perhaps the most striking result of this study was the 0.15 extent to which the genes regulated by RalA and/or 0.1 RalB and the genes found to be differentially expressed

Normalized RLU in bladder tumors vs normal tissue overlapped. Nearly 0.05 14% (84/607) of the probe sets differentially expressed in 0 control RalA I RalB II RalA/B siRNA

c Figure 3 Ral regulates RREB-1 in UMUC-3cells. ( a) UMUC-3 FLAG-RalA stably expressing an RREB-1-luciferase reporter responding to RalA G12V activated H-Ras expression. Data shown are the averages FLAG-RalB of duplicate experiments, normalized to cell number using RalB CyQUANT, with error bars indicating standard deviation (*, Po0.05, t-test). Inset, western blot showing H-Ras expression. (b) RREB-1-driven luciferase activity in UMUC-3cells transfected 0.12 serum with Ral siRNA. Data shown are the averages of duplicate serum starved * * experiments, normalized to cell number using CyQUANT, with 0.1 error bars indicating standard deviation (*, Po0.05, t-test). Inset, western blot indicating RalA and RalB protein levels in siRNA- 0.08 transfected cells from representative parallel transfection. (c) 0.06 RREB-1-driven luciferase activity in UMUC-3cells transfected with G23V, activated RalA or RalB. Data shown are the averages 0.04 of duplicate experiments, normalized to cell number using CyQUANT, with error bars indicating standard deviation (*, Normalized RLU 0.02 Po0.05, t-test). Inset, western blot indicating FLAG-tagged RalA and RalB protein levels in transfected cells from representative 0 parallel transfection. RREB, Ras-responsive element-binding FLAG ActRalA ActRalB protein; siRNA, short interfering RNA.

Oncogene RalA and RalB functional expression proﬁling in human bladder cancer G Oxford et al 7150 a 4.5 * 4 3.5 3 * UMUC-6 2.5 * * KU-7 * 2 HeLa * * * 1.5 PC-3 1 * * 0.5 * Relative RREB-luciferase actvity 0 FLAG actH-Ras actRalA actRalB

b FLAG ActH-RasFLAG ActH-Ras ActRalA ActRalB

UMUC-6 HA-H-Ras FLAG-Ral

KU-7 HA-H-Ras FLAG-Ral

HeLa HA-H-Ras FLAG-Ral

PC-3 HA-H-Ras FLAG-Ral

Figure 4 Ral Regulates RREB-1 in multiple cancer cell lines. (a) Luciferase activity of four human cancer cell lines transiently cotransfected with RREB-1-luciferase reporter and G12V-activated H-Ras or G23V-activated RalA or RalB. Data shown are the averages of duplicate experiments, normalized both to cell number using CyQUANT and to FLAG control within each cell line, with error bars indicating standard deviation (*, Po0.05, t-test vs FLAG control). (b) Western blots showing expression levels of HA- tagged G12V-activated H-Ras and FLAG-tagged G23V-activated RalA and RalB in representative transfections from (a). RREB, Ras-responsive element-binding protein

Table 4A Number of expected and observed probe sets that were bladder cancer strongly suggests that transcriptional regulated, in response to RalA and/or RalB knockdown as well as regulation by RalA and RalB contributes significantly differentially expressed in human bladder cancer vs normal urothelium to the development of human bladder cancer. However, Expecteda,b Observeda,c P-valued because higher RalA and RalB activation states are thought to contribute to cancer formation and progres- RalA regulated genes 6/225 25/225 8 Â 10À15 RalB regulated genes 7/256 17/256 0.00011 sion (Lim et al., 2005, 2006), the siRNA-mediated RalA/B regulated genes 20/732 64/732 3 Â 10À24 approach we took may have uncovered different Ral- regulated genes than those regulated by endogenously Sixty-five bladder tumors and 20 normal bladder urothelium samples activated Ral in cancer cells. With Ral depletion were profiled by HG-U133A chips and compared to Ral GTPase triggering both increases and decreases in gene expres- a siRNA treated UMUC3cells (see text for details). Overlap with probe sion, dissecting out the complex relationships between sets differentially expressed in human bladder cancer vs normal urothelium (607/22283). bExpected ¼ 607/22283multiplied by the total RalA/B activation status and specific gene expression number of probe sets significantly regulated by RalA and/or RalB changes in bladder cancer progression will require depletion via siRNA. cObserved ¼ (number of probe sets significantly further experiments. Meanwhile, among the genes which regulated by RalA and/or RalB depletion via siRNA which are also are differentially expressed in both bladder cancer and differentially expressed as a function of human bladder cancer vs normal urothelium) divided by (number of probe sets significantly regulated by Ral depletion are vascular endothelial regulated by RalA and/or RalB depletion via siRNA). dCalculated by growth factor (VEGF), which plays a key role in tumor w2 test. angiogenesis (Carmeliet, 2005), and CD24, which we previously reported to be regulated by RalA/B (Smith human bladder cancer were also affected by Ral et al., 2006). The regulation of both VEGF and CD24 knockdown. Overrepresentation of Ral-regulated genes by RalA/B may represent an opportunity for therapeu- in a set of genes differentially expressed in human tic intervention.

Oncogene RalA and RalB functional expression proﬁling in human bladder cancer G Oxford et al 7151 Materials and methods Computational ascertainment of regulatory relationships analysis Cell culture and transfection Identifying putative transcription factors associated with gene Cell culture and transfections of UMUC-3, UMUC-6, KU-7, expression changes triggered by Ral siRNA transfection was PC-3and HeLa cancer cells with siRNA duplexes speciﬁc for carried out using CARRIE Web Service (Haverty et al., 2004) RalA and/or RalB or complementary DNA constructs are (http://zlab.bu.edu/CarrieServer/html/). More detail is pro- described in detail in the Supplementary Materials. vided in the Supplementary Materials.

Gene expression profiling and gene ontology analysis Significance of Ral-regulated genes in human bladder cancer UMUC-3cells were transfected with siRNA duplexes target- By combining previously published gene expression profiling ing RalA, RalB, RalA and RalB together or firefly luciferase data sets of human bladder cancer (Dyrskjot et al., 2004; (GL2). At 72 h after siRNA transfection, lysates were Nicholson et al., 2004) and our Ral siRNA data, we evaluated harvested in RLT buffer (Qiagen, Valencia, CA, USA) and the number of probe sets that were altered as a function of Ral RNA isolation performed as described previously. Gene modulation and differentially expressed in human bladder expression profiling of cells transfected with Ral siRNA was cancer using standard statistics. More detail is provided in the carried out using the HG-U133A Array (Affymetrix Inc., Supplementary Materials. Santa Clara, CA, USA). More detail is provided in the Supplementary Materials. Acknowledgements Western blotting and reporter assays Western blotting for RalA, RalB and FLAG was carried out This work was supported by Medical Scientist Training as described previously (Oxford et al., 2005). Luciferase Program training grant T32GM007267 to SCS, Cancer reporter assays for NF-kB and RREB-1 transcriptional Training Grant CA009109-29 to SCS, and CA075115 and activity were performed with Luciferase Assay System from PO1CA104106 to DT. Promega (Madison, WI, USA) as described previously (Bigler et al., 2003), using CyQUANT (Molecular Probes, Eugene, Conflict of interest OR, USA) to normalize results by cell numbers. These authors state no conflict of interest.

References

Awasthi S, Cheng J, Singhal SS, Saini MK, Pandya U, Pikula Hamad NM, Elconin JH, Karnoub AE, Bai W, Rich JN, S et al. (2000). Novel function of human RLIP76: ATP- Abraham RT et al. (2002). Distinct requirements for Ras dependent transport of glutathione conjugates and doxo- oncogenesis in human versus mouse cells. Genes Dev 16: rubicin. Biochemistry 39: 9327–9334. 2045–2057. Awasthi S, Singhal SS, Sharma R, Zimniak P, Awasthi YC. Haverty PM, Frith MC, Weng Z. (2004). CARRIE web (2003). Transport of glutathione conjugates and chemothera- service: automated transcriptional regulatory network peutic drugs by RLIP76 (RALBP1): a novel link between G- inference and interactive analysis. Nucleic Acids Res 32: protein and tyrosine kinase signaling and drug resistance. W213–W216. Int J Cancer 106: 635–646. Henry DO, Moskalenko SA, Kaur KJ, Fu M, Pestell RG, Bigler D, Gulding KM, Dann R, Sheabar FZ, Conaway MR, Camonis JH et al. (2000). Ral GTPases contribute to Theodorescu D. (2003). Gene proﬁling and promoter regulation of cyclin D1 through activation of NF-kappaB. reporter assays: novel tools for comparing the biological Mol Cell Biol 20: 8084–8092. effects of botanical extracts on human prostate cancer cells Jebar AH, Hurst CD, Tomlinson DC, Johnston C, Taylor CF, and understanding their mechanisms of action. Oncogene 22: Knowles MA. (2005). FGFR3and Ras gene mutations are 1261–1272. mutually exclusive genetic events in urothelial cell carcino- Camonis JH, White MA. (2005). Ral GTPases: corrupting the ma. Oncogene 24: 5218–5225. exocyst in cancer cells. Trends Cell Biol 15: 327–332. Jullien-Flores V, Dorseuil O, Romero F, Letourneur F, Carmeliet P. (2005). VEGF as a key mediator of angiogenesis Saragosti S, Berger R et al. (1995). Bridging Ral GTPase in cancer. Oncology 69(Suppl 3): 4–10. to Rho pathways. RLIP76, a Ral effector with CDC42/ Chien Y, White MA. (2003). RAL GTPases are linchpin Rac GTPase-activating protein activity. J Biol Chem 270: modulators of human tumour-cell proliferation and survi- 22473–22477. val. EMBO Rep 4: 800–806. Lebreton S, Boissel L, Iouzalen N, Moreau J. (2004). RLIP Dyrskjot L, Kruhoffer M, Thykjaer T, Marcussen N, Jensen mediates downstream signalling from RalB to the actin JL, Moller K et al. (2004). Gene expression in the urinary cytoskeleton during Xenopus early development. Mech Dev bladder: a common carcinoma in situ gene expression 121: 1481–1494. signature exists disregarding histopathological classiﬁcation. Lim KH, Baines AT, Fiordalisi JJ, Shipitsin M, Feig LA, Cox Cancer Res 64: 4040–4048. AD et al. (2005). Activation of RalA is critical for Ras-induced Feig LA. (2003). Ral-GTPases: approaching their 15 min of tumorigenesis of human cells. Cancer Cell 7: 533–545. fame. Trends Cell Biol 13: 419–425. Lim KH, O’Hayer K, Adam SJ, Kendall SD, Campbell PM, Gildea JJ, Harding MA, Seraj MJ, Gulding KM, Theodorescu Der CJ et al. (2006). Divergent roles for RalA and RalB in D. (2002). The role of Ral A in epidermal growth malignant growth of human pancreatic carcinoma cells. factor receptor-regulated cell motility. Cancer Res 62: Curr Biol 16: 2385–2394. 982–985. Moskalenko S, Tong C, Rosse C, Mirey G, Formstecher E, Grossman HB, Wedemeyer G, Ren L, Wilson GN, Cox B. Daviet L et al. (2003). Ral GTPases regulate exocyst (1986). Improved growth of human urothelial carcinoma cell assembly through dual subunit interactions. J Biol Chem cultures. J Urol 136: 953–959. 278: 51743–51748.

Oncogene RalA and RalB functional expression profiling in human bladder cancer G Oxford et al 7152 Nicholson BE, Frierson HF, Conaway MR, Seraj JM, protein, is involved in the differentiation response to Ras in Harding MA, Hampton GM et al. (2004). Profiling the human medullary thyroid carcinomas. Mol Cell Biol 16: evolution of human metastatic bladder cancer. Cancer Res 5335–5345. 64: 7813–7821. Ward Y, Wang W, Woodhouse E, Linnoila I, Liotta L, Kelly Ohta Y, Suzuki N, Nakamura S, Hartwig JH, Stossel TP. K. (2001). Signal pathways which promote invasion and (1999). The small GTPase RalA targets filamin to induce metastasis: critical and distinct contributions of extracellular filopodia. Proc Natl Acad Sci USA 96: 2122–2128. signal-regulated kinase and Ral-specific guanine exchange Oxford G, Owens CR, Titus BJ, Foreman TL, Herlevsen MC, factor pathways. Mol Cell Biol 21: 5958–5969. Smith SC et al. (2005). RalA and RalB: antagonistic White MA, Nicolette C, Minden A, Polverino A, Van Aelst L, relatives in cancer cell migration. Cancer Res 65: 7111–7120. Karin M et al. (1995). Multiple Ras functions can contribute Shi Y, Sawada J, Sui G, Affar el B, Whetstine JR, Lan F et al. to mammalian cell transformation. Cell 80: 533–541. (2003). Coordinated histone modifications mediated by a Wolthuis RM, Zwartkruis F, Moen TC, Bos JL. (1998). Ras- CtBP co-repressor complex. Nature 422: 735–738. dependent activation of the small GTPase Ral. Curr Biol 8: Shipitsin M, Feig LA. (2004). RalA but not RalB enhances 471–474. polarized delivery of membrane proteins to the basolateral Wu X, Obata T, Khan Q, Highshaw RA, De Vere White R, surface of epithelial cells. Mol Cell Biol 24: 5746–5756. Sweeney C. (2004). The phosphatidylinositol-3kinase Smith SC, Oxford G, Wu Z, Nitz MD, Conaway M, Frierson pathway regulates bladder cancer cell invasion. BJU Int HF et al. (2006). The metastasis-associated gene CD24 is 93: 143–150. regulated by Ral GTPase and is a mediator of cell Zhang S, Qian X, Redman C, Bliskovski V, Ramsay ES, proliferation and survival in human cancer. Cancer Res 66: Lowy DR et al. (2003). p16 INK4a gene promoter variation 1917–1922. and differential binding of a repressor, the ras-responsive Thiagalingam A, De Bustros A, Borges M, Jasti R, Compton zinc-finger transcription factor, RREB. Oncogene 22: D, Diamond L et al. (1996). RREB-1, a novel zinc finger 2285–2295.

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

Oncogene