Interleukin-1R–Associated Kinase 2 Is a Novel Modulator of the Transforming Growth Factor Β Signaling Cascade

Published OnlineFirst March 23, 2010; DOI: 10.1158/1541-7786.MCR-09-0386 Published Online First on March 23, 2010

Molecular Signaling and Regulation Cancer Research Interleukin-1R–Associated Kinase 2 Is a Novel Modulator of the Transforming Growth Factor β Signaling Cascade

Jasper Mullenders, Armida W.M. Fabius, Miranda M.W. van Dongen, Hendrik J. Kuiken, Roderick L. Beijersbergen, and René Bernards

Abstract The transforming growth factor β (TGFβ) pathway orchestrates an extensive transcriptional program that is important for many processes in the cell. For example, TGFβ regulates cell cycle, migration, and epithelial-to- mesenchymal transition. The TGFβ pathway has a dual role in cancer: it is involved in early-stage tumor suppression but also contributes to tumor progression by promoting invasion. To identify the novel genes involved in TGFβ pathway signaling, we have performed a functional genetic loss-of-function screen. We screened a small interfering RNA library targeting 700 kinases and kinase-related genes in a TGFβ-responsive reporter assay. Several genes were identified that upon knockdown could repress the reporter signal; among these are the two cellular receptors for TGFβ. In addition to these two known components of the TGFβ pathway, several genes were identified that were previously not linked to the TGFβ signaling. Knockdown of one of these genes, the IRAK2 kinase, resulted not only in an impaired TGFβ target gene response but also in a reduction of the nuclear accumulation and phosphorylation of SMAD2. In addition, suppression of interleukin-1R–associated kinase 2 expression led to a partial override of a TGFβ-induced cell cycle arrest. Our data show that interleukin- 1R–associated kinase 2 is a novel and critical component of TGFβ signaling. Mol Cancer Res; 8(4); OF1–12. ©2010 AACR.

Introduction In the nucleus, the activated SMADs can form a complex with SMAD4. This activated SMAD complex recruits The transforming growth factor β (TGFβ) pathway is an transcriptional cofactors that assist in regulating the tran- important signaling pathway that regulates many different scription of genes whose promoters contain SMAD bind- processes such as cell cycle, epithelial-to-mesenchymal tran- ing DNA elements (3). sition, migration, and angiogenesis (1, 2). Stimulation of The TGFβ pathway plays a dual role in cancer patho- the TGFβ pathway is initiated by binding of the TGFβ cyto- genesis. In normal epithelium and early-stage tumors, the kine to the TGFβ receptors (TGFBR1 and TGFBR2). This TGFβ pathway was reported to have an inhibitory effect leads to TGFβ receptor complex formation, which results on cell proliferation (4). This is achieved mainly through in the phosphorylation of TGFBR1 by TGFBR2. Subse- the induction of a G1 cell cycle arrest through the upregu- quently, SMAD2 and SMAD3 are phosphorylated by the lation of the CDK inhibitors CDKN1A (encoding p21) TGFBR1, which leads to their translocation to the nucleus. and CDKN2B (encoding p15), and the downregulation of the MYC proto-oncogene (5-7). However, cancer cells can become insensitive to this proliferation control by loss of expression of the TGFβ receptors or SMADs. In addition, it has been observed that some tumors have specifi- Authors' Affiliation: Division of Molecular Carcinogenesis, Center for β Biomedical Genetics and Cancer Genomics Center, The Netherlands cally inactivated the cytostatic response to TGF while Cancer Institute, Amsterdam, the Netherlands retaining normal SMADs and receptors. In these tumors, β Note: Supplementary data for this article are available at Molecular the TGF pathway subsequently promotes cancer progres- Cancer Research Online (http://mcr.aacrjournals.org/). sion through the induction of epithelial-to-mesenchymal J. Mullenders and A.W.M. Fabius contributed equally to this work. transition, angiogenesis, and evasion of immune surveil- Current address for M.M.W. van Dongen: Division of Pathology, The lance (1, 2). Furthermore, it was recently shown that pa- Netherlands Cancer Institute, Amsterdam, the Netherlands. tients carrying tumors with an activated TGFβ pathway Corresponding Authors: René Bernards and Roderick L. Beijersbergen, have a worse prognosis than patients carrying tumors with- Division of Molecular Carcinogenesis, Center for Biomedical Genetics β and Cancer Genomics Center, The Netherlands Cancer Institute, Plesman- out activation of the TGF pathway (8, 9). For this reason, laan 121, 1066 CX Amsterdam, the Netherlands. Phone: 31-20-512-1952; different anticancer strategies are developed to inhibit Fax: 31-20-512-1954. E-mail: [email protected] and [email protected] the TGFβ pathway. Clinical trials have been done with doi: 10.1158/1541-7786.MCR-09-0386 TGFβ antisense oligonucleotides, TGFβ antibodies, and ©2010 American Association for Cancer Research. a TGFBR1 small-molecule inhibitor (10).

www.aacrjournals.org OF1

Mullenders et al.

Several approaches have been taken to identify new play- 14 h. Cell titer blue reagent (Promega) was used to deter- β ers in the TGF pathway. For example, protein-protein in- mine the cell viability. After 1 h of incubation, 560EX/ teraction screens have been used to extend the knowledge 590EM fluorescence was measured with a plate reader (En- about the TGFβ pathway (11, 12). In addition, a functional vision multilabel reader 2101, Perkin-Elmer). Subsequently, genetic screen to identify new players in the mammalian the medium was aspirated with a robot (STAR Liquid TGFβ network has been done by Levy and colleagues (13). Handling Workstation, Hamilton) and luminescence was They describe the screening of a small interfering RNA measured with the same plate reader using Steady-Glo (siRNA) library targeting the ubiquitin E3 ligase gene Luciferase (Promega). family in a cell line with an integrated TGFβ-responsive Arkadia reporter. Through this approach, they identified as Statistical Analysis a positive regulator of the pathway. For the analysis of the human kinase siRNA screen, the We describe here a functional genetic screen that aimed following statistical analysis was done. The cell viability at the identification of novel kinases, which function in the counts were used to calculate a viability score and siRNA β TGF pathway. We identify here an unexpected new ki- pools that reduce cell viability by >25% were omitted from β nase that contributes to TGF signaling. further validation rounds. The luciferase counts were cor- rected for cell viability giving the normalized luciferase counts (NORM LUC). Subsequently, the NORM LUC Materials and Methods counts were log2 transformed to get a normal distribution. The transformed NORM LUC counts are used to calculate Cell Lines and Culture Conditions the Z-score (difference of the sample luciferase measure- U2OS, HaCaT, and PC3 cells were cultured in DMEM ment and the average luciferase of the population divided supplemented with 10% FCS, penicillin, streptomycin, and by the SD of the population) per plate to be able to compare glutamine. All cells were cultured at 37°C in 5% CO2. the different plates with each other. Further validation was done on the siRNA pools with Z-scores of <−2 or >2. Generation of Stable Reporter Cell Line For the follow-up experiments, we used one-tailed Stu- U2OS cells were transfected using calcium phosphate dent's t tests to calculate the P value of the sample compared β with the TGF -responsive reporter pGL3-CAGA12-Luc with the control siRNA. and a puromycin-resistant plasmid pBabe-puro in a 10:1 ratio. Next, the cells were sparsely seeded and selected with Validation of Single siRNAs 2 μg/mL puromycin. Subsequently, colonies were picked The four separate siRNAs and the siRNA pool were tested and tested for the TGFβ-dependent induction of luciferase in the U2OS-CAGA cell line. As a control, we used a signal. A single clone was selected based on the level of lu- siRNA-targeting green fluorescent protein (GFP). All single ciferase expression and the inducibility of the luciferase sig- siRNA validation experiments were done at least thrice using nal by TGFβ. quadruplicate transfection and one representative experiment is shown. siRNA Transfection Cells were reverse transfected with siRNAs according to Immunohistochemistry to Quantify SMAD 2/3 manufacturers instructions (Dharmacon; day 1). Dharma- Nuclear Accumulation fect 1 (Dharmacon) was used for U2OS cells and Dharma- PC3 cells were cultured in 96- or 384-well plates. Form- fect 2 (Dharmacon) was used for HaCaT and PC3 cells. aldehyde was used for the fixation of the cells and we per- At day 2, penicillin and streptomycin was added. TGFβ meabilized the cells with 0.2% Triton X-100. Subsequently, (200 pmol/L, R&D Systems; human platelet derived) was samples were blocked with 5% bovine serum albumin in added overnight (U2OS) or for 24 h (HaCaT). For PC3 0.05% Tween 20 in PBS. A SMAD2/3-specific (BD Trans- cells, TGFβ was added for 8 h to assay target gene activa- duction Laboratories, 610842) primary antibody was used tion or 1 h for the SMAD2/3 nuclear accumulation assay. and subsequently a mouse ALEXA 488–conjugated (Alexa Fluor 488, Invitrogen) secondary antibody. The cells were ′ Human Kinase siRNA Screen counterstained with 4 ,6-diamidino-2-phenylindole to de- Cells with stably integrated TGFβ-responsive reporter fine the position of the nuclei. Images were acquired with were reverse transfected in 384-well plates in triplicate with a high-content imager (BD Pathway Bioimager 855, BD the human kinome library of Dharmacon. Dharmafect 1 biosciences) using a ×20 objective (Olympus). CellProfiler (0.1 μL per 384-well) was used as transfection reagent and software (14) was used to quantify the nuclear/cytoplasmic the siRNA pools were transfected at a concentration of SMAD2/3 ratio of all individual cells per well. 50 nmol/L. Three thousand cells were seeded per well with a cell dispenser (Wellmate, Matrix). The transfection was Western Blotting done in absence of antibiotics. Some 18 h after the transfec- Cell lysates were separated using 4% to 12% Bis-Tris gels tion, we added the penicillin and streptomycin. Forty-eight (Nupage, Invitrogen). Proteins were transferred to a polyvi- hours after transfection, cells were stimulated with TGFβ for nylidene difluoride membrane and incubated with primary

OF2 Mol Cancer Res; 8(4) April 2010 Molecular Cancer Research

siRNA Screen Identifies New TGFβ Pathway Component

antibody as indicated. Primary antibodies were detected screen was done in the presence of TGFβ.Toidentify using a secondary horseradish peroxidase–conjugated anti- siRNAs that modulate the reporter, we measured both lucif- body. Antibodies used for these studies were as follows: erase activity and cell viability. The cell viability was measured Hsp90 (Santa Cruz H-114, sc-7947), SP1 (Santa Cruz to exclude siRNAs that cause a decrease in cell viability as PEP 2, sc-59), SMAD2 [Cell Signaling (L16D3) #3103], these might be identified as false-positive repressors in the SMAD3 [Cell Signaling (C67H9) #9523], p-SMAD2 screen. Data were normalized and a hit list was generated ser465/467 (Cell Signaling #3101), and p-SMAD3 by calculating a Z-score for every tested siRNA pool (Fig. 1B). ser423/425 [Cell Signaling (C25A9) #9520]. Only siRNA pools that produced a Z-score of >2 (20 genes) or lower than −2 (17 genes) were selected for fur- Quantitative PCR ther analysis (Supplementary Table S1). Among the genes Total RNA was isolated using Trizol (Invitrogen). From targeted by siRNA pools that repressed the reporter were the total RNA, cDNA was generated using SuperScript II the positive controls: TGFBR1 and TGFBR2. In addition, (Invitrogen) using random primers (Invitrogen). cDNA we identified the siRNA pool against ALK1 (ACVRL1) as was diluted and the quantitative real-time PCR (QRT) re- an activator of the TGFβ-responsive reporter. ALK1 was action was done using FAST SYBR green (Invitrogen) with previously identified as a TGFβ superfamily receptor type I specific primers (Supplementary Table S2). All quantitative protein that can form a complex with TGFBR2 and direct- PCRs were run in parallel with RPL13 to control for input ly antagonizes TGFβ transcriptional activation mediated by cDNA. The QRTreactions were done using a fast Real-time TGFBR1 (17, 18). The fact that the screen was able to iden- PCR system (AB7500, Applied Biosystems). tify these genes that are part of the canonical TGFβ signaling network supports the notion that at least some of the other Cell-Cycle Analysis genes in the hit list are putative players in the TGFβ pathway. For fluorescence-activated cell sorting analysis, siRNA- As a first step to validate our screening results, we retested transfected HaCaT cells that were treated with TGFβ the same siRNA pools that were identified by the screen to for 24 h were fixed, stained, and assayed as previously verify that they could indeed reproduce the phenotype from described (15). the primary screen. Therefore, we transfected 37 siRNA pools into the U2OS-CAGA cell line and measured the Results luciferase signal. For 36 of the 37 siRNA pools, we could reproduce the phenotype as measured in the primary screen, The Screening of a TGFβ Reporter Cell Line indicating that the results from the initial screen with the with a Human Kinome siRNA Library siRNA pools are robust (Fig. 1C and D). To screen siRNA libraries in high-throughput format One of the major drawbacks of using RNAi as a screening for novel modulators of TGFβ signaling, we generated tool is that some observed phenotypes can potentially be an U2OS osteosarcoma cell line with a stably integrated caused by “off-target” effects(19,20).Therefore,weper- β pCAGA12-Luciferase TGF -responsive reporter (this cell formed a second round of validation for a subset of our hits. lines was named U2OS-CAGA). This reporter has been We selected hits based on the position in the primary hit list. described to be primarily responsive to SMAD3/4-dependent In addition, protein-protein interaction databases were used transcription (16). To test the responsiveness of the inte- to preferentially select hits that interact with known TGFβ grated reporter in the U2OS-CAGA cell line, we transfected pathway members. Sixteen hits (eight siRNA pools that siRNAs targeting a positive and a negative regulator could activate the reporter and eight that repressed) were of TGFβ signaling. As expected, ablation of the positive reg- selected for a second round of validation. In this second ulator TGFBR2 by RNAi led to complete abrogation of both round of validation, the four siRNAs of each siRNA pool basal and TGFβ-induced reporter activity (Fig. 1A). More- were transfected separately to test if they could reproduce over, knockdown of a negative regulator of the TGFβ signal- the originally observed phenotype. For 9 of the 16 hits test- ing cascade, SKIL, caused the opposite effect, showing ed, we identified two individual siRNAs that recapitulated increased reporter activity. It is worth noting that in the ab- the phenotype that was obtained with the siRNA pools sence of exogenously added TGFβ, knockdown of SKIL still (repressor siRNAs and b activator siRNAs; Fig. 2A). In causes enhanced signal of the TGFβ-responsive reporter addition, all siRNAs that could either repress or activate (Fig. 1A). We also observed a robust TGFβ-dependent in- the reporter could also significantly repress the intended tar- duction of the CAGA12 reporter in cells transfected with get mRNA (Fig. 2C and D). Therefore, we consider the siRNAs targeting GFP (Fig. 1A). genes targeted by these siRNAs to be “on target” (21). These initial experiments showed that the integrated TGFβ reporter in the U2OS-CAGA cell line behaved as ex- Effect of Knockdown of On-Target Hits pected. Therefore, we used this cell line to screen a siRNA on Endogenous TGFβ Signaling library targeting 700 kinases and kinase-related proteins. Because the reporter system used here is only an indirect The siRNA library that we used consists of siRNA pools; measurement of TGFβ pathway activity, we set out to test if each gene is targeted by four separate siRNAs. The the validated hits could also affect endogenous TGFβ target U2OS-CAGA cells were transfected in triplicate with gene expression. Many different genes are directly regulated siRNA pools of the human kinome library and the entire by TGFβ through the SMADs; among these genes are

www.aacrjournals.org Mol Cancer Res; 8(4) April 2010 OF3

Mullenders et al.

FIGURE 1. Screen with TGFβ-responsive reporter identifies modulators of the TGFβ pathway. U2OS-CAGA cells harboring a stable TGFβ-responsive luciferase reporter were transfected with siRNAs against GFP, TGFBR2, or SKIL and treated with 200 pmol/L TGFβ (+) for 14 h or left nontreated (−). Using a paired one-tailed t test, we found that the P value for the treated samples (TGFBR2 and SKIL) was below 0.05. A, results from the kinase siRNA screen; Z-scores were calculated and plotted in ascending order. Boxed areas, the siRNA pools selected for follow-up, which repress or activate the TGFβ-responsive reporter (B). Seventeen siRNA pools that repress the reporter (C) and 20 siRNAs that activate the reporter (D) were tested in a follow-up experiment. The normalized luciferase counts from the screen (white columns) and follow-up (gray columns) are plotted. P values for the primary and secondary screen can be found in supplementary Table 1.

CDKN1A, SMAD7, PAI-1, and c-MYC. As a model for tion of the TGFβ reporter (four in total) as these genes are endogenous TGFβ signaling, we used the prostate cancer potential novel targets for therapy in cancer. These four hits, cell line PC3. In these cells, transcription of CDKN1A, together with TGFBR1 as a control, were tested for their SMAD7, and PAI-1 is increased by TGFβ treatment, where- ability to modulate endogenous TGFβ target gene expres- as c-MYC is transcriptionally repressed after addition of sion. Indeed, all hits tested impaired TGFβ-induced re- TGFβ. Because we identified a relative large number (nine) sponses on these endogenous target genes (Fig. 3A-D). of hits that proved to be on target, we decided to proceed Furthermore, we checked if the siRNAs also effectively sup- with the genes whose downregulation impaired the activa- pressed their intended target. In all cases, we observed an

OF4 Mol Cancer Res; 8(4) April 2010 Molecular Cancer Research

siRNA Screen Identifies New TGFβ Pathway Component

80% or more reduction of mRNA levels of the targets of the fected with siRNAs targeting the on-target hits to determine transfected siRNAs (Fig. 3E). the nuclear/cytoplasmic ratio of SMAD2/3 in the presence of TGFβ. A clear reduction in SMAD2/3 nuclear accumu- Identified Hits Are Involved in SMAD2/3 lation was measured in cells transfected with siRNAs against Nuclear Accumulation TGFBR1. In addition, we measured a lower TGFβ-induced β TGF target gene activation is a downstream event in the SMAD2/3 nuclear/cytoplasmic ratio for cells transfected β TGF signaling cascade, which is preceded by the transloca- with siRNA pools against IRAK2, PKIB,andSTK22B tion of SMAD2/3 complexes from the cytoplasm to the (Fig. 4B). This suggests that like the TGFBR1, interleukin- nucleus. For this reason, we tested if the genes that could 1R–associated kinase (IRAK2), PKIB, and STK22B function β repress TGF target gene activation also have impaired upstream or at the level of SMAD2/3 nuclear accumula- SMAD2/3 nuclear accumulation. As a control, we measured tion in the TGFβ pathway. the effect of the knockdown of TGFBR1 on TGFβ-induced SMAD2/3 nuclear accumulation by immunohistochemis- IRAK2 Is a Genuine Player in the TGFβ try. We quantified the effect by determining the ratio of Signaling Cascade SMAD2/3 in the nucleus versus the cytoplasm using the For the remainder of this study, we limited ourselves to CellProfiler software (Fig. 4A; ref. 14). PC3 cells were trans- the investigation of the role of IRAK2 in TGFβ signaling.

FIGURE 2. Nine hits from the screen are on target. Individual siRNAs from the siRNA pool were tested in the U2OS-CAGA cell line. A GFP siRNA was used as negative control. Two individual siRNAs and the pool that could either repress (A) or activate (B) the CAGA reporter are shown. Knockdown for these individual siRNAs was determined by quantitative PCR (C and D). All values in both the reporter and quantitative PCR experiments are significantly (P < 0.05) different compared with the GFP siRNA.

www.aacrjournals.org Mol Cancer Res; 8(4) April 2010 OF5

Mullenders et al.

FIGURE 3. Endogenous TGFβ target gene regulation. The on target siRNA pools that repress the reporter were transfected in PC3 cells and target gene expression was measured in the absence or presence of TGFβ for 8 h. Quantitative PCR was performed for c-MYC (A), PAI-1 (B), SMAD7 (C), and CDKN1A (D). The mRNA levels of the target genes were determined by quantitative PCR and normalized to the reference gene RPL13. Knockdown of the intended targets by the transfection of the used siRNAs was also measured by comparing to control-transfected (siGFP) cells; the percentage of remaining mRNA is plotted (D). Shown here is a representative experiment that has been done thrice. All on-target hits can significantly (P < 0.05) repress TGFβ-induced changes in gene expression.

OF6 Mol Cancer Res; 8(4) April 2010 Molecular Cancer Research

siRNA Screen Identifies New TGFβ Pathway Component

Therefore, we decided to perform additional experiments human keratinocytes. As expected, a TGFBR2-targeting to find out if IRAK2 can functionally modulate the siRNA pool can prevent the proliferative arrest, retaining TGFβ pathway. In addition, we set out to investigate the amount of S-phase cells. We also observed that knock- the exact mechanism of inhibition of the TGFβ pathway down of IRAK2 can also partly prevent this proliferative by IRAK2 knockdown. Previously, IRAK2 was described arrest (Fig. 5A). In addition, we checked the efficacy of to play a role in Toll-like receptor signaling (22). As knockdown by the used siRNAs in HaCat cells. As mea- shown before, the knockdown of IRAK2 impaired the sured by quantitative PCR, the siRNAs conferred 80% TGFβ target gene activation of c-MYC, PAI-1, and 84% reduction of TGFBR2 and IRAK2 mRNA le- CDKN1A, and SMAD7 (Fig. 3). At least three of these vels, respectively (Fig. 5B and C). genes are involved in cell cycle regulation (c-MYC, PAI-1, and CDKN1A) and it has been shown that at least p21 IRAK2 Knockdown Affects SMAD2, but not SMAD3, and PAI-1 are required for a TGFβ-induced proliferative Accumulation in the Nucleus arrest (23, 24). For this reason, we setup an experiment As shown earlier, knockdown of the identified hits has a to test if IRAK2 is also required for aTGFβ-induced cell clear effect on the TGFβ-induced nuclear accumulation of β cycle arrest. TGF is known to induce a G1 cell cycle ar- SMAD2/3. However, it must be noted that the antibody rest and consequently fewer cells will enter the DNA rep- used in the immunofluorescence assay (Fig. 4) detects both lication phase (S phase). This arrest can be clearly the SMAD2 and SMAD3 proteins. This can potentially lead measured in the HaCat cell line, which is derived from to the misinterpretation of the results as the individual

FIGURE 4. Effect of knockdown of the hits on SMAD2/3 nuclear accumulation. PC3 cells were transfected with siRNAs against GFP or TGFBR1, and 48 h, later treated with TGFβ (+) for 1 h or left nontreated (−). Immunohistochemistry was done with an antibody specific for SMAD2/3. A, PC3 cells were transfected with four siRNA pools identified in the screen together with negative (GFP) and positive control (TGFBR1). Forty-eight hours after transfection, cells were treated with TGFβ for 1 h. Cells were fixed and stained with an anti-SMAD2/3 antibody and cells were counterstained with 4′,6-diamidino-2-phenylindole to determine the position of the nuclei. The ratio of nuclear and cytoplasmic SMAD2/3 is quantified with CellProfiler (B). Genes whose knockdown significantly (P < 0.05) represses SMAD2/3 nuclear accumulation in the presence of TGFβ are shown (*).

www.aacrjournals.org Mol Cancer Res; 8(4) April 2010 OF7

Mullenders et al.

FIGURE 5. IRAK2 is a genuine regulator of TGFβ signaling. HaCaT cells were transfected with GFP, TGFBR2, or IRAK2 siRNAs (siRNA pool) and cells were treated with TGFβ (+ TGFβ) for 24 h or left nontreated (−). The cells were fixed and stained with propidium iodide. Cell cycle profiles were measured and the percentage of cells in S phase are denoted. Knockdown of TGFBR2 (B) and IRAK2 (C) was determined in HaCat cells by measuring mRNA levels with quantitative PCR. The repression of IRAK2 and TGFBR2 by their respective siRNAs compared with the GFP siRNA is statistically significant (P < 0.05).

contributions of SMAD2 and SMAD3 cannot be deter- served after treatment with TGFβ.Strikingly,cellstransfected mined. For this reason, we measured the effect of IRAK2 with IRAK2 siRNAs showed a normal nuclear accumulation knockdown on SMAD2 and SMAD3 individually. There- of SMAD3, but an impaired nuclear accumulation of fore, we transfected PC3 cells with either control or IRAK2 SMAD2 pointing to a specific role for IRAK2 in the activa- siRNAs and, after incubation, treated the cells with TGFβ tion and therefore nuclear accumulation of SMAD2. for 1 hour. Nuclear and cytoplasmic extracts were prepared and the lysates were analyzed by Western blot using antibo- IRAK2 Knockdown Impairs SMAD2 dies that can specifically detect either SMAD2 or SMAD3 Phosphorylation (Fig. 6A). In control-transfected cells (siRNA GFP), regular SMAD2/3 nuclear accumulation is preceded by the phos- nuclear accumulation of SMAD2 and SMAD3 can be ob- phorylation of the SMADs by the TGFβ receptors (25-27).

OF8 Mol Cancer Res; 8(4) April 2010 Molecular Cancer Research

siRNA Screen Identifies New TGFβ Pathway Component

For this reason, we measured the level of phosphorylation of knockdown impairs phosphorylation and translocation of both SMAD2 and SMAD3 upon IRAK2 knockdown. PC3 SMAD2 but not SMAD3 to the nucleus. Therefore, cells were transfected with siRNAs targeting GFP and we assessed the effect of SMAD2 knockdown on the IRAK2, and incubated for 48 hours. Transfected cells were TGFβ-responsive reporter and on endogenous TGFβ target subsequently treated with TGFβ for 30, 60, and 120 min- genes. First, we tested if knockdown of SMAD2 just like utes. Protein lysates were prepared and analyzed by Western SMAD3 represses the TGFβ-responsive reporter. We used blot. A clear reduction in phosphorylated SMAD2 was ob- the U2OS-CAGA cells and transfected control (GFP), served at all time points. This supports the notion that SMAD2 or SMAD3 siRNAs (Fig. 7A). From this experi- IRAK2 is specifically required for SMAD2 activation by ment, we conclude that a SMAD2 siRNA is indeed able TGFβ (Fig. 6B). to repress the TGFβ-responsive reporter. Subsequently, we tested if the used SMAD2 and 3 siRNAs are specific for their SMAD2 Is also Required for TGFβ Target intended target. As shown in the Western blot analysis Gene Induction (Fig. 7B), SMAD2 siRNA do not repress SMAD3 protein It has previously been postulated that SMAD3 and not and vice versa. This indicates that the used reagents are SMAD2 is responsible for most TGFβ-induced changes indeed SMAD selective. Furthermore, we could show that in transcription. However, we have shown that IRAK2 endogenous TGFβ target gene induction is also partly impaired in PC3 cells transfected with SMAD2 siRNAs (Fig. 7C-F). To control the efficacy of the knockdown of SMAD2 and SMAD3 by the used siRNAs, we performed quantitative PCR. The result from this experiment shows that both used siRNAs give >80% knockdown of SMAD2 and SMAD3 mRNA levels (Fig. 7G and H). Overall, the observed repression of the TGFβ-responsive reporter and TGFβ-dependent target genes by knockdown of SMAD2 supports the finding that IRAK2 exerts its function on the TGFβ pathway through an effect on SMAD2.

Discussion

In this study, we describe the identification of new modulators of the TGFβ pathway through the screening of a siRNA library targeting kinases and kinase-related genes. Data analysis of the screen revealed a relatively large number of genes that could either repress or activate the TGFβ-responsive reporter. Somewhat surprising was the fact that the data distribution is exponential instead of linear. However, this was also observed in two previously done reporter-based RNAi screens (13, 28). Some RNAi screening efforts have reported high levels of off-target effects caused by nonspecific suppression of mRNAs (29, 30). Therefore, it is satisfying to find that ∼56% (9 of 16 genes) of the hits tested could be classified as on target (21). Observations made in reporter assays are not necessarily reflecting endogenous pathway activation. Therefore, we tested if the five on target hits that repressed the reporter also repress endogenous target gene induction. This experiment was done in a different cell line than the original U2OS- CAGA to exclude the cell type specificity of the identified FIGURE 6. IRAK2 knockdown impairs SMAD2 and SMAD3 nuclear hits. For all five on target hits, impaired activation of endog- accumulation and phosphorylation. PC3 cells were transfected with enous target genes was observed. The fact that the newly GFP and IRAK2 siRNAs (siRNA pool), and cells were treated with identified hits can regulate the TGFβ-dependent transcrip- TGFβ (+) for 1 h or left nontreated (−). Nuclear and cytoplasmic fractions β were immunoblotted for HSP90; cytosolic loading control SP1, nuclear tion of certain TGF targets suggests that these genes indeed loading control, and antibodies specific for SMAD2 and SMAD3 (A). have a role in the TGFβ pathway. To study the effect of IRAK2 knockdown on SMAD2 and SMAD3 The TGFβ/SMAD signaling pathway can be regulated phosphorylation, cells were transfected with IRAK2 siRNAs and at many different levels: cofactors that regulate SMAD- subsequently treated with TGFβ for various times. Whole-cell lysates were immunoblotted for HSP90 (loading control), total SMAD2 and dependent transcription, nucleoporin proteins assist in SMAD3 protein, and with phospo-SMAD2 (Ser465/476) and SMAD nuclear accumulation, and of course the phosphor- phospho-SMAD3 (Ser423/425) antibodies (B). ylation by the TGFβ receptors. Strikingly, for three of four

www.aacrjournals.org Mol Cancer Res; 8(4) April 2010 OF9

Mullenders et al.

FIGURE 7. SMAD2 and SMAD3 are both required for TGFβ target gene regulation. U2OS-CAGA cells were transfected with siRNAs against GFP, SMAD2, or SMAD3 and treated with 200 pmol/L TGFβ (+) for 14 h or left nontreated (−). Both SMAD2 and SMAD3 could significantly (P < 0.05) repress the TGFβ-dependent induction of the CAGA-Luc reporter compared the GFP siRNA. (A). To control the specificity of the used SMAD2 and SMAD3 siRNAs, we transfected PC3 cells with the respective siRNAs and monitored protein levels of SMAD2 and SMAD3 by immunoblotting (B). SMAD2 and SMAD3 are both essential for TGFβ target gene induction. PC3 cells were transfected with the indicated siRNAs and treated with TGFβ for 8 h. RNA was isolated and quantitative PCR was used to measure the induction of TGFβ target genes: c-Myc (C), PAI-1 (D), SMAD7 (E), and CDKN1A (F). In all cases, the repression of the target genes is significant (P < 0.05). To control the efficiency and specificity of the obtained knockdown, RNA from the PC3 cells was used for a real-time PCR with primers specific for SMAD2 (G) and SMAD3 (H). A paired t test showed that the knockdown of either SMAD2 or SMAD3 compared with GFP siRNA was significant (P < 0.05).

OF10 Mol Cancer Res; 8(4) April 2010 Molecular Cancer Research

siRNA Screen Identifies New TGFβ Pathway Component

of the newly identified modulators of TGFβ signaling nase domain (35, 36). One of the first publications on (IRAK2, PKIB, and STK22B), we observed a defective IRAK2 described its function as a downstream mediator SMAD2/3 nuclear accumulation. This points into the di- of the IL-1R in a complex with TRAF6 and MyD88 (37). rection that these genes function upstream of SMAD nu- It was only recently unveiled that IRAK2 does have an active clear accumulation. kinase domain and has a functional role in Toll-like receptor signaling (22, 38). Our observation that IRAK2 ablation IRAK2 Is Specifically Required for SMAD2 Activation leads to an impaired TGFβ response suggests a possible in- We initially showed that IRAK2 knockdown impairs the teraction between TGFβ and IL-1/TLR signaling. This is TGFβ-dependent changes in the transcription of all tested supported by various previously done studies. For example, TGFβ target genes. Further analysis showed that IRAK2 the fact that IRAK2 was found in a luminescence-based pro- seems specifically required for the TGFβ-dependent phos- teomics screen to interact with both SMAD2 and SMURF1 phorylation and subsequent nuclear accumulation of (12). In addition, it was recently shown that TRAF6, which SMAD2. We subsequently showed that SMAD2 knock- interacts with IRAK2, mediates the TGFβ-dependent down impairs the activation of a TGFβ-responsive reporter activation of c-Jun-NH2-kinase and p38. This effect was and the induction of TGFβ-dependent target genes. This reported to be SMAD independent, proven by the fact observation was unexpected because the TGFβ-responsive that TRAF6 knockdown does not lead to activation of a reporter has been described to be dependent on SMAD3/ TGFβ-responsive reporter (39). Finally, it was shown that 4-dependent transcription. However, this conclusion was the stimulation of cells with IL-1β leads in many cases to based on experiments, which showed that SMAD3 or phosphorylation of SMAD2 and subsequent target gene SMAD4 overexpression activate this TGFβ-responsive re- activation (40). Taken together, our and other observations porter, whereas SMAD2 overexpression did not (16). Sever- indicate the existence of cross-talk between two important al studies have reported that SMAD3 is responsible for the cellular signaling pathways involved in immunity and majority of the TGFβ-induced phenotypes and that other processes. SMAD2 is often dispensable (31-33). More recently, it was reported that under some conditions, SMAD2 indeed Disclosure of Potential Conflicts of Interest plays a critical role in transmitting TGFβ signals from the cell membrane to the nucleus (34). Some apparent interplay R. Bernards, commercial research grant, Schering Plough. between the two receptor activated SMADs could also be Acknowledgments seen when we knocked down SMAD2 and IRAK2. We observed an increase of SMAD3 protein levels after IRAK2 We thank C. Lieftink and B. Evers for the technical assistance, P. ten Dijke and J. (Fig. 6A and B) or SMAD2 knockdown (Fig. 7B). Further- Seoane for the reagents and the helpful discussion, T. Kuilman for providing the sequences for the RPL13 and Actin B QRT primers, and A. Pfauth for the help more, SMAD3 mRNA is also elevated after SMAD2 knock- with Cellquest. down (Fig. 7H), suggesting that a feedback mechanism is existent to compensate for the loss of SMAD2 or IRAK2. Grant Support More experiments are required to fully understand the mechanism of the loss of SMAD2 phosphorylation in cells EU 6th framework integrated project INTACT, The Netherlands Genomics with knockdown of IRAK2. Initiative, the Centre of Biomedical Genetics, the Cancer Genomics Centre, and Schering Plough (Oss, the Netherlands). The costs of publication of this article were defrayed in part by the payment of page IRAK2 and SMAD2: Connecting IL-1 and charges. This article must therefore be hereby marked advertisement in accordance TGFβ Signaling with 18 U.S.C. Section 1734 solely to indicate this fact. IRAK2 is part of a family of four IRAKs; despite its name, Received 08/24/2009; revised 02/05/2010; accepted 02/21/2010; published it was predicted that IRAK2 does not contain an active ki- OnlineFirst 03/23/2010.

References 1. Siegel PM, Massague J. Cytostatic and apoptotic actions of TGF-β 8. Bruna A, Darken RS, Rojo F, et al. High TGFβ-Smad activity confers in homeostasis and cancer. Nat Rev Cancer 2003;3:807–21. poor prognosis in glioma patients and promotes cell proliferation 2. Massague J. TGFβ in Cancer. Cell 2008;134:215–30. depending on the methylation of the PDGF-B gene. Cancer Cell 3. Massague J, Seoane J, Wotton D. Smad transcription factors. Genes 2007;11:147–60. Dev 2005;19:2783–810. 9. Penuelas S, Anido J, Pieto-Sanchez RM, et al. TGF-β increases glioma- 4. Massague J, Gomis RR. The logic of TGFβ signaling. FEBS Lett initiating cell self-renewal through the induction of LIF in human 2006;580:2811–20. glioblastoma. Cancer Cell 2009;15:315–27. 5. Datto MB, Li Y, Panus JF, et al. Transforming growth factor β induces 10. Iyer S, Wang ZG, Akhtari M, Zhao W, Seth P. Targeting TGFβ the cyclin-dependent kinase inhibitor p21 through a p53-independent signaling for cancer therapy. Cancer Biol Ther 2005;4:261–6. mechanism. Proc Natl Acad Sci U S A 1995;92:5545–9. 11. Colland F, Jacq X, Trouplin V, et al. Functional proteomics mapping 6. Hannon GJ, Beach D. p15INK4B is a potential effector of TGF-β- of a human signaling pathway. Genome Res 2004;14:1324–32. induced cell cycle arrest. Nature 1994;371:257–61. 12. Barrios-Rodiles M, Brown KR, Ozdamar B, et al. High-throughput 7. Seoane J, Pouponnot P, Staller P, et al. TGFβ influences Myc, Miz-1 and mapping of a dynamic signaling network in mammalian cells. Smad to control the CDK inhibitor p15INK4b. Nat Cell Biol 2001;3:400–8. Science 2005;307:1621–5.

www.aacrjournals.org Mol Cancer Res; 8(4) April 2010 OF11

Mullenders et al.

13. Levy L, Howell M, Das D, et al. Arkadia activates Smad3/Smad4- 27. Souchelnytskyi S, Tamaki K, Engstrom U, et al. Phosphorylation of dependent transcription by triggering signal-induced SnoN degrada- Ser465 and Ser467 in the C terminus of Smad2 mediates interaction tion. Mol Cell Biol 2007;27:6068–83. with Smad4 and is required for transforming growth factor-β signaling. 14. Carpenter AE, Jones TR, Lamprecht MR, et al. CellProfiler: image J Biol Chem 1997;272:28107–15. analysis software for identifying and quantifying cell phenotypes. 28. Firestein R, Bass AJ, Kim SY, et al. CDK8 is a colorectal cancer on- Genome Biol 2006;7:R100. cogene that regulates β-catenin activity. Nature 2008;455:547–51. 15. Berns K, Hijmans EM, Mullenders J, et al. A large-scale RNAi screen 29. Ma Y, Creanga A, Lum L, Beachy PA. Prevalence of off-target effects in human cells identifies new components of the p53 pathway. Nature in Drosophila RNA interference screens. Nature 2006;443:359–63. 2004;428:431–7. 30. Tang W, Dodge M, Gundapaneni P, et al. A genome-wide RNAi 16. Dennler S, Itoh S, Vivien D, et al. Direct binding of Smad3 and Smad4 screen for Wnt/β-catenin pathway components identifies unexpected to critical TGF β-inducible elements in the promoter of human plas- roles for TCF transcription factors in cancer. Proc Natl Acad Sci U S A minogen activator inhibitor-type 1 gene. EMBO J 1998;17:3091–100. 2008;105:9697–702. 17. Oh SP, Seki T, Goss KA, et al. Activin receptor-like kinase 1 modulates 31. Yang YC, Piek E, Zavadil J, et al. Hierarchical model of gene regula- transforming growth factor-β 1 signaling in the regulation of angio- tion by transforming growth factor β. Proc Natl Acad Sci U S A 2003; genesis. Proc Natl Acad Sci U S A 2000;97:2626–31. 100:10269–74. 18. Goumans MJ, Valdimarsdottir G, Itoh S, et al. Activin receptor-like 32. Piek E, Ju WJ, Heyer J, et al. Functional characterization of trans- kinase (ALK)1 is an antagonistic mediator of lateral TGFβ/ALK5 sig- forming growth factor β signaling in Smad2- and Smad3-deficient naling. Mol Cell 2003;12:817–28. fibroblasts. J Biol Chem 2001;276:19945–53. 19. Perrimon N, Mathey-Prevot B. Matter arising: off-targets and 33. Brown KA, Pietenpol JA, Moses HL. A tale of two proteins: differen- genome-scale RNAi screens in Drosophila. Fly (Austin) 2007;1:1–5. tial roles and regulation of Smad2 and Smad3 in TGF-β signaling. 20. Svoboda P. Off-targeting and other non-specific effects of RNAi J Cell Biochem 2007;101:9–33. experiments in mammalian cells. Curr Opin Mol Ther 2007;9:248–57. 34. Yang J, Wahdan-Alaswad R, Danielpour D. Critical role of Smad2 21. Echeverri CJ, Beachy PA, Baum B, et al. Minimizing the risk of in tumor suppression and transforming growth factor-β-induced reporting false positives in large-scale RNAi screens. Nat Methods apoptosis of prostate epithelial cells. Cancer Res 2009;69:2185–90. 2006;3:777–9. 35. Janssens S, Beyaert R. Functional diversity and regulation of 22. Kawagoe T, Sato S, Matsushita K, et al. Sequential control of Toll-like different interleukin-1 receptor-associated kinase (IRAK) family receptor-dependent responses by IRAK1 and IRAK2. Nat Immunol members. Mol Cell 2003;11:293–302. 2008;9:684–91. 36. Hanks SK, Hunter T. Protein kinases 6. The eukaryotic protein kinase 23. Bachman KE, Blair BG, Brenner K, et al. p21(WAF1/CIP1) mediates superfamily: kinase (catalytic) domain structure and classification. the growth response to TGF-β in human epithelial cells. Cancer Biol FASEB J 1995;9:576–96. Ther 2004;3:221–5. 37. Muzio M, Ni J, Peng F, Dixit VM. IRAK (Pelle) family member IRAK-2 24. Kortlever RM, Nijwening JH, Bernards R. Transforming growth and MyD88 as proximal mediators of IL-1 signaling. Science 1997; factor-β requires its target plasminogen activator inhibitor-1 for 278:1612–5. cytostatic activity. J Biol Chem 2008;283:24308–13. 38. Keating SE, Maloney GM, Moran EM, Bowie AG. IRAK-2 participates 25. Abdollah S, Macias-Silva M, Tsukazaki T, et al. TβRI phosphorylation in multiple toll-like receptor signaling pathways to NFκB via activa- of Smad2 on Ser465 and Ser467 is required for Smad2-4 complex tion of TRAF6 ubiquitination. J Biol Chem 2007;282:33435–43. formation and signaling. J Biol Chem 1997;272:27678–85. 39. Yamashita M, Fatyol K, Jin C, et al. TRAF6 mediates Smad-independent 26. Liu Q, Huang SS, Huang JS. Function of the type V transforming activation of JNK and p38 by TGF-β. Mol Cell 2008;31:918–24. growth factor β receptor in transforming growth factor β-induced 40. Lu T, Tian L, Han Y, Vogelbaum M, Stark GR. Dose-dependent cross- growth inhibition of mink lung epithelial cells. J Biol Chem 1997; talk between the transforming growth factor-β and interleukin-1 272:18891–5. signaling pathways. Proc Natl Acad Sci U S A 2007;104:4365–70.

OF12 Mol Cancer Res; 8(4) April 2010 Molecular Cancer Research

Interleukin-1R−Associated Kinase 2 Is a Novel Modulator of the Transforming Growth Factor β Signaling Cascade

Jasper Mullenders, Armida W.M. Fabius, Miranda M.W. van Dongen, et al.

Mol Cancer Res Published OnlineFirst March 23, 2010.

Updated version Access the most recent version of this article at: doi:10.1158/1541-7786.MCR-09-0386

Supplementary Access the most recent supplemental material at: Material http://mcr.aacrjournals.org/content/suppl/2010/03/23/1541-7786.MCR-09-0386.DC1

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://mcr.aacrjournals.org/content/early/2010/03/20/1541-7786.MCR-09-0386. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.