DDX5 is a positive regulator of oncogenic NOTCH1 signaling in T cell acute lymphoblastic leukemia

Abstract

Notch signaling is a highly conserved cell-cell communication pathway regulating normal development and tissue homeostasis. Aberrant Notch signaling represents an important oncogenic mechanism for T cell acute lymphoblastic leukemia (T-ALL), an aggressive subset of the most common malignant childhood cancer ALL. Therefore, understanding the molecular regulation of Notch signaling is critical to identifying new approaches to block aberrant Notch oncogenic activity. The family of three MAML transcriptional co-activators is crucial for Notch signaling activation. The prototypic member MAML1 is the major co-activator that regulates Notch oncogenic activities in leukemic cells. However, the molecular basis underlying MAML1 co-activator function that contributes to Notch signaling remains unclear. In this study, we performed proteomic studies and identified DDX5, an ATP-dependent DEAD-box RNA helicase, as a component of the MAML1 protein complex. DDX5 interacts with MAML1 in vitro and in vivo, and is associated with the endogenous NOTCH1 transcription activation complex in human T-ALL leukemic cells. Lentivirus-mediated shRNA knockdown of DDX5 resulted in decreased expression of Notch target genes, reduced cell proliferation, and increased apoptosis in cultured human leukemic cells with constitutive activation of Notch signaling. Also, DDX5 depletion inhibited the growth of human leukemia xenograft in nude mice. Moreover, DDX5 is highly expressed in primary human T-ALL leukemic cells based on the analyses of Oncomine and GEO databases and Immunohistochemical staining. Our overall findings revealed a critical role of DDX5 in promoting efficient Notch-mediated transcription in leukemic cells, suggesting that DDX5 might be critical for NOTCH1-mediated T-ALL pathogenesis and thus is a potential new target for modulating the Notch signaling in leukemia.

Introduction

Notch signaling is a highly conserved developmental signaling pathway, regulating cell proliferation, differentiation, and survival.¹ Deregulated Notch signaling is linked to many epithelial cancers and hematological malignancies.^2,3 In particular, aberrant Notch signaling represents an important oncogenic mechanism for T cell acute lymphoblastic leukemia (T-ALL), an aggressive subset of the most common malignant childhood cancer ALL. More than 60% of T-ALL cases harbor activated NOTCH1 mutations.^4,5 Sustained high levels of NOTCH1 signaling cause leukemic cell transformation and are required for the maintenance of the leukemic phenotype⁶; thus the Notch pathway is a promising, highly specific molecular therapeutic target for T-ALL. Moreover, the Notch pathway also is an important target for solid cancers including breast cancer, lung cancer, pancreatic cancer, brain tumors and melanoma.⁷ Consequently, manipulation of Notch signaling has a great promise as a major new therapeutic strategy for treating the large array of Notch-hyperactive cancers.

Canonical Notch signaling is initiated by ligand-receptor interactions between adjacent cells.⁸ Ligand binding results in proteolytic cleavages of Notch receptors and subsequent release of the intracellular domain of Notch (ICN) from the cell membrane. ICN then enters the nucleus and forms a complex with the transcription factor CSL which recruits the transcriptional co-activators including Mastermind-like (MAML) and other cofactors such as histone acetyltransferase, resulting in transcriptional activation. Our current understanding of molecular mechanisms regulating Notch-mediated transcription remains incomplete. The elucidation of new critical regulators of the Notch signaling pathway will lead to the identification of therapeutic targets that modify deregulated Notch activity in cancers.

Our group and others have previously identified the family of three MAML transcriptional co-activators (MAML1, 2, 3) that are essential components of the Notch transcription activation complex.^9–11 MAML1 is a major co-activator that regulates NOTCH1 oncogenic activities in leukemic cells, and is essential for maintaining leukemic cell growth and survival (unpublished). Hence, our data suggest that the manipulation of MAML1 expression or functional activities will affect leukemia initiation and progression. Currently, the molecular basis underlying the MAML1 co-activation function remains very limited.¹² We hypothesize that cellular factors associated with MAML1 contribute to the MAML1-co-activation function in enhancing Notch-mediated transcription by molecular mechanisms, for instance, through interacting with chromatin remodeling/modifying cofactors or promoting formation or functions (transcription or elongation) of the pre-initiation complex.

Therefore, to gain insights into the molecular basis underlying MAML1 function in regulating Notch signaling, we undertook the task of identifying cellular factors that interact with MAML1 and characterizing their roles in Notch signaling regulation and Notch-mediated T cell leukemia. In this study, we revealed that DDX5 (also known as p68), an ATP-dependent RNA helicase,^13,14 is a component of the MAML1 protein complex and is associated with the Notch transcription complex. Given a role for DDX5 as a transcriptional modulator,^15,16 we reasoned that DDX5 might be a critical regulator of Notch-mediated transcription and consequently of oncogenic Notch signaling. Our findings in this study demonstrated a biochemical link of DDX5 with the Notch transcription complex through its interaction with MAML1, and revealed a critical role of DDX5 for efficient Notch-mediated transcription in leukemic cells and potentially in other human cancers in which the Notch pathway is active. Therefore, our study suggests that DDX5 is critical for Notch-mediated T-ALL pathogenesis and has the potential as a drug target in leukemia.

Results

DDX5 is identified as a MAML1-interacting protein in a proteomic study

We recently found that MAML1 is a major transcriptional co-activator contributing to Notch signaling in human T-ALL leukemic cells that contain NOTCH1-activating mutations, as depletion of MAML1 expression resulted in a significant decrease in Notch target gene expression (unpublished), similar to the effects of γ-secretase inhibitor treatment that blocks Notch receptor processing.¹⁷ We hypothesized that MAML1 functions by recruiting specific co-regulators that activate transcription of Notch target genes. MAML1 was previously shown to interact with p300/CBP, CDK8 and GSK3β^18–21 that exert positive or negative effects on Notch transactivation. However, these interactions cannot account for all MAML1 co-activation function. To systematically identify co-regulators that are important for MAML1 transcriptional co-activation, we isolated the endogenous MAML1 protein complex from activating NOTCH1 mutations-containing leukemic cells (KOPT-K1) by immunoprecipitation using rabbit antibodies generated from the MAML1 peptides. Mass spectrometric analysis was then performed to identify proteins in the MAML1 protein complex. We found that MAML1 is associated with known interacting proteins including NOTCH1 and the Notch pathway transcription factor CSL, indicative of the presence of the core transcriptional complex, NOTCH1/CSL/MAML1. We also identified novel potential MAML1-interacting proteins. In this study, we focused on one of the MAML1-interacting protein candidates, the ATP-dependent RNA helicase DDX5, since DDX5 has been functionally linked to transcriptional co-activation and many oncogenic events.^15,16

MAML1 interacts with DDX5 in vitro and in vivo

We validated the interaction between MAML1 and DDX5 using a series of assays. First, GST pull-down assays were performed to determine whether DDX5 specifically interacts with MAML1 in vitro. Bacterially expressed GST or GST-MAML1 fusion proteins were immobilized on glutathione-sepharose, which were used as baits for cellular lysates expressing FLAG-tagged DDX5. We found that GST tagged full-length (FL) or 1-302 aa of MAML1, but not GST, specifically pulled down FLAG-tagged DDX5 (Figure 1a), indicating a specific interaction between the N terminal region of MAML1 and DDX5 in vitro. Second, co-immunoprecipitation assays were performed to determine the in vivo interaction of FLAG-tagged MAML1 and Myc-tagged DDX5. 293T cells were transfected with plasmids encoding FLAG-tagged MAML1 or Myc-tagged DDX5 individually or together. MAML1 was immunoprecipitated with anti-FLAG antibodies and then any associated Myc-tagged DDX5 was detected by immunoblotting with anti-Myc antibodies. We found that MAML1 readily co-precipitated with DDX5 when overexpressed in the cells (Figure 1b). Third, we used co-immunoprecipitation assays to determine the interaction between MAML1 and DDX5 at the endogenous protein levels. Here, HeLa cell lysates were prepared and endogenous MAML1 immunoprecipitation using rabbit anti-MAML1 antibodies was performed. We found that DDX5 co-immunoprecipitated with MAML1 endogenously in HeLa cells (Figure 1c). To further confirm the interaction between MAML1 and DDX5 in a more physiological condition, we performed mammalian two-hybrid assays. The wild-type (wt) DDX5 and enzyme-inactive K144R mutant (mutation in ATP binding domain, mut) constructs were expressed as fusion proteins with a GAL4 DNA binding domain, and MAML1 was expressed as a fusion protein with the activation domain (AD) in U2OS cells. The interaction between DDX5 and MAML1 was quantified by the activation of a luciferase reporter containing GAL4 binding sites in the promoter. We found that the luciferase activity from cells expressing both proteins was significantly increased compared to the control, indicating that MAML1 and DDX5 interact with each other in vivo (Figure 1d). Moreover, the enzyme-inactive mutant of DDX5 appeared not to interact with MAML1 (Figure 1d). Overall, the above data indicate a specific interaction between MAML1 and DDX5 in vitro and in vivo.

Figure 1. MAML1 specifically interacts with DDX5 in vitro and in vivo.

(a) MAML1 interacts with DDX5 by GST pull-down assays. GST-MAML1 FL and 1-302aa fusion proteins were induced expressed in BL21 bacterial cells, purified, and then incubated with cellular lysates from 293T cells transfected with FLAG-DDX5. The pull-down proteins were analyzed by Western blotting with anti-FLAG antibodies. * indicates the protein bands of GST or GST-MAML1 fusions. (b) MAML1 co-immunoprecipitates with DDX5. 293T cells were transfected with FLAG-MAML1 and/or Myc-DDX5 plasmids, and cell lysates were collected 48 hours after transfection for IP and Western blot analysis. (c) MAML1 and DDX5 interact at the endogenous protein levels in HeLa cells. Whole cell lysates collected from HeLa cells were used for immunoprecipitation with anti-MAML1 antibodies. The MAML1 immunoprecipitates were analyzed by Western blotting. (d) MAML1 interacts with wild-type (wt), but not enzyme-inactive version (mut), of DDX5 by mammalian two-hybrid assays. U2OS cells were co-transfected with pBIND-DDX5 and different amounts of MAML1 plasmids. The cell lysates were collected for luciferase assays 48 hours after transfection. Data were shown as average values from three independent experiments. **p<0.01.

DDX5 is associated with the Notch transcription activation complex in leukemic cells

It was reported that more than 60% of T-ALL patients carry activating NOTCH1 mutations, which result in the formation of active NOTCH1/CSL/MAML1 transcriptional complex and subsequent expression of Notch target genes, contributing to leukemic phenotype.^5,6 To explore the potential functions of DDX5 in T-ALLs, we first surveyed the expression levels of leukemic cells. By Western blot analysis, we found that DDX5 is expressed in a variety of T-ALL leukemic cell lines and DDX5 expression levels appear to be higher in a majority of T-ALL cells in comparison with three AML (acute myeloid leukemia) cell lines we tested (Figure 2a).

Figure 2. DDX5 is associated with the Notch transcription activation complex in leukemic cells.

(a) DDX5 expression levels in a panel of AML and T-ALL leukemia cell lines were determined by Western blot analysis. (b) MAML1 interacts with DDX5 in T-ALL cells. KOPT-K1 cells were treated with GSI (1 μM ComE) for 24 hours and then lysates were collected for IP with anti-MAML1 antibodies and blotted with indicated antibodies. (c) NOTCH1 interacts with DDX5 in KOPT-K1 cells. Protein lysates of KOPT-K1 cells were used for IP with an anti-NOTCH1 antibody raised against the intracellular domain of the NOTCH1 receptor, and then blotted with an anti-DDX5 antibody. (d) DDX5 is associated with the core Notch transcriptional activation complex (i.e. NOTCH1/CSL/MAML1). IP was performed using KOPT-K1 cellular lysates with an anti-DDX5 antibody following by Western blotting with indicated antibodies.

Based on the evidence that MAML1 interacts with DDX5 in a protein complex and that DDX5 functions in transcriptional co-activation, DDX5 might have a role in mediating MAML1-coactivator function and thereby modulating Notch signaling in T-ALL leukemia. Therefore, we next determined whether DDX5 is associated with the NOTCH1 transcription activation complex. First, we determined whether MAML1 interacts with DDX5 in the presence or the absence of active NOTCH1 signaling. Here, activating NOTCH1 mutation-containing KOPT-K1 leukemic cells were treated with the vehicle control dimethyl sulfoxide (DMSO) or γ-secretase inhibitor (GSI) that interferes with NOTCH1 receptor processing and hence blocks Notch signaling, and then cell lysates were prepared for endogenous MAML1 immunoprecipitation using anti-MAML1 antibodies. We found that GSI treatment greatly reduced cleaved NOTCH1 levels (Figure 2b), indicating effective blockade of NOTCH1 signaling. In consistence with these data, cleaved NOTCH1 and CSL levels were significantly reduced in the MAML1 immunoprecipitates (Figure 2b and not shown), indicating a reduced level of active Notch transcription complex. Importantly, we found that the levels of DDX5 co-immunoprecipitated with MAML1 remained similar in both the presence and absence of active Notch signaling (Figure 2b), suggesting that MAML1 interacted with DDX5 irrespective of cellular Notch signal status. Next, we performed NOTCH1 (ICN1) immunoprecipitation using antibodies against the cytoplasmic domain of NOTCH1, and detected any possible DDX5 co-immunoprecipitated with NOTCH1. We found that DDX5 is readily detectable in NOTCH1 immunoprecipitates (Figure 2c). The reverse immunoprecipitation using anti-DDX5 antibodies also supported an interaction between DDX5 and the NOTCH1(ICN1)/MAML1/CSL transcription complex (Figure 2d). Since DDX5 is an RNA binding protein, we further determined whether there is a role for RNA in the interaction of DDX5 and MAML1. KOPT-K1 protein lysates were first treated with RNase/DNase to remove the nuclear acids before being subjected to immunoprecipitation. Our data showed that MAML1 was able to pull down DDX5 protein after RNase/DNase treatment, indicating that the interaction between MAML1 and DDX5 is independent of nuclear acids (Supplementary Figure 1). These data indicate that DDX5 interacts with MAML1 independent of cellular Notch signaling status and is recruited to the NOTCH1 (ICN1) transcription complex, hence suggesting a role for DDX5 in modulating Notch signaling in Notch-active leukemic cells.

DDX5 enhances Notch-mediated transcription and DDX5 depletion reduces expression of Notch signature genes in leukemic cells

To evaluate a potential role of DDX5 in Notch-mediated transcription, we first determined whether DDX5 promotes NOTCH1 (ICN1)-induced transcription. Here, U2OS cells were transfected with a Notch-responsive promoter reporter (pCSL-luc containing four copies of CSL-binding sites in the promoter) and activated NOTCH1 (ICN1) in the presence and absence of wild-type or enzymatically inactive forms of DDX5 (Figure 3a). We found that DDX5 enhanced NOTCH1 (ICN1)-induced reporter activity and such enhancement is dependent on the intact DDX5 enzymatic activity, as the mutation (K144R) renders inactive enzyme failed to show enhanced activity (Figure 3a). These data suggest that DDX5 activation of Notch-mediated transcription is dependent on DDX5 enzyme activity. Moreover, chromatin immunoprecipitation (ChIP) assay revealed that DDX5 was enriched on the CSL-binding promoter region of the Notch target gene HES1 and that the level of DDX5 protein associated with HES1 promoter was reduced in DDX5 knockdown cells (Figure 3b), which further supported a function of DDX5 in the regulation of Notch-mediated transcription.

Figure 3. DDX5 enhances Notch-mediated transcription and its depletion reduces expression levels of Notch target genes in NOTCH1 mutated KOPT-K1 cells.

(a) DDX5 enhances Notch mediated transcription. U2OS cells were co-transfected with intracellular domain of NOTCH1 (ICN1), DDX5 (wildtype and mutant) together with the NOTCH responsive promoter reporter pCSL-Luc and Renilla luciferase plasmid (for normalization of transfection). Luciferase activities were assayed about 48 hours after transfection. Data were shown as average from three independent experiments. **p<0.01; ***p<0.001. (b) DDX5 directly binds to NOTCH target HES1 promoter. ChIP assay was performed in KOPT-K1 cells using DDX5 antibody or control IgG. The upper panel shows the location of primers on HES1 promoter. (c) DDX5 is effectively knocked down in KOPT-K1 cells. KOPT-K1 cells were infected with lentiviruses that contain two independent shRNAs that target KD DDX5, and the control viruses expressing shRNA target luciferase gene. Transduced cells were assayed for DDX5 expression by Western blotting analysis (upper panel) and real-time RT-PCR (lower panel). (d) Real-time RT-PCR assays showed that DDX5 depletion reduces expression levels of NOTCH target genes. (b–d) the Real-time PCR data were shown as average from three independent experiments.

Next, we investigated whether DDX5 loss-of-function impairs Notch signaling in leukemic cells. We utilized a pLKO.1-based lentiviral shRNA set containing five U6 promoter-regulated shRNA targeting DDX5, and identified two forms of shRNA (shDDX5-3 and shDDX5-4) that were effective to knock down DDX5 in HeLa cells (not shown). We then infected activating NOTCH1 mutation-bearing cells, KOPT-K1, with these two forms of DDX5 shRNA (shDDX5-3, and -4) or luciferase shRNA (shLuc) as controls, and collected RNA and protein samples for analysis. We found that there was about 70% reduction in the DDX5 transcript level by real-time RT-PCR assays and more than 90% of DDX5 protein reduction though Western blot analysis (Figure 3c). We subsequently compared the levels of endogenous Notch target genes in DDX5 knockdown and control cells by real-time RT-PCR. We found that DDX5 depletion by two forms of shRNA resulted in decreased expression levels of Notch target genes including HES1, HEY1, MYC and DTX1 (Figure 3d). In contrast, knockdown of DDX5 failed to significantly affect Notch target gene expression in SUPT13 T-ALL cells that have wild-type NOTCH1 gene and are insensitive to Notch signaling inhibition (Supplementary Figure 2). These data indicate that DDX5 is essential for efficient Notch-mediated transcription in Notch-active leukemic cells.

DDX5 regulates leukemic cell proliferation and survival

Since activated NOTCH1 signaling is essential for leukemic cell growth and survival and DDX5 modulates Notch signaling, we predicted that DDX5 loss-of-function and subsequent reduced Notch signaling will inhibit leukemic cell proliferation and survival. Therefore, we knocked down DDX5 expression in a series of leukemic cells by lentiviral infection on two consecutive days followed by puromycin selection for 2 days. At day 6 post-infection (D0), cells were set up for cell proliferation, cell cycle, and apoptosis assays while cell lysates were made simultaneously to determine the extent of DDX5 knockdown. We were able to knock down DDX5 protein levels by about 70% and 90% in several cell lines including KOPT-K1, HPB-ALL, MOLT4 and Jurkat (Figure 3c and not shown). We found that DDX5 knockdown led to a decrease in cell growth in these Notch-overactive leukemic cells, but not in the Notch insensitive SUPT13 cells (Figure 4a and Supplementary Figure 2). To study the mechanisms underlying cell growth suppression caused by DDX5 loss-of-function, we first investigated the impact of DDX5 knockdown on cell cycle profile using BrdU/7-AAD staining followed by FACS analysis. We found that DDX5 depletion resulted in inhibition of G1-S-phase transition (Figure 4b, Supplementary Figure 3). Next, we investigated whether DDX5 loss-of-function affects cell survival by various apoptotic assays. We found that DDX5 knockdown led to enhanced apoptotic activities by measuring Caspase-Glo 3/7 activity (Figure 4c), cleaved Caspase 3 level by Western blot analysis (Figure 4d) and Annexin V/PI staining (Figure 4e, Supplementary Figure 4). These data indicates that DDX5 depletion-induced cell growth suppression could be due to both cell cycle arrest and induction of apoptosis. Combined with the previous data, our findings indicate that DDX5 modulates cell proliferation and survival in leukemic cells at least in part through regulating Notch signaling.

Figure 4. DDX5 knockdown results in a reduction in T-ALL cell proliferation and survival.

KOPT-K1 cells were infected with lentiviruses expressing shRNA against DDX5 or control shLuc on two consecutive days, and then selected with puromycin for 2 days. At 6 days after the first infection (considered D0), cell growth, cell cycle and survival assays (a–e) were performed. (a) DDX5 knockdown reduces T-ALL leukemic cell proliferation. Control and DDX5-knockdown leukemic cells were cultured under the same conditions (a total of 2 ml at 1×10⁵ cells/ml) and cell numbers were counted at day 2 (D2) and D4. Data were shown as average values from three independent experiments. **p<0.01. (b) Cell cycle distribution of DDX5 knockdown and control cells were determined by BrdU/7-AAD straining followed by FACS analysis. (c) DDX5 knockdown caused an increase in Caspase 3/7 activities in leukemic cells. About 1×10⁴ cells were used for the quantification of Caspase 3/7 activities. Data were shown as average from three independent experiments. **p<0.01; ***p<0.001. (d) DDX5 knockdown caused an increased level of cleaved Caspase 3 level by Western blot analysis. (e) DDX5 knockdown caused an increased percentage of apoptotic cells in KOPT-K1 cells determined by Annexin V-PI staining followed by FACS analysis.

DDX5 depletion results in reduced T-ALL cell growth in vivo

To test the function of DDX5 in Notch-overactive T-ALL cell growth in vivo, we generated HPB-ALL cells expressing the luciferase gene (HPB-ALL-luc) by infecting the HPB-ALL cells with retroviruses that co-express GFP and luciferase under a bicistronic transcript. FACS sorting for GFP+ cells was performed to enrich transduced cells, which were further validated for luciferase expression by luciferase assays. HPB-ALL-luc cells were then infected with lentiviruses that expressed the shRNA against DDX5 or lentiviruses generated from the vector control. At about 72 hours after first infection, cell lysates were collected for the confirmation of DDX5 knockdown by Western blotting (Figure 5a). At the same day, an equivalent number of the viable DDX5-depleted and control cells (based on Trypan blue assays and luciferase assays shown in Supplementary Figure 5) were subcutaneously injected into nude mice. The growth of tumor xenografts was then monitored by measuring tumor volume and luciferase intensities at different time points after leukemia cell injection (Figure 5b). The tumors in the mice injected with the control cells grew and reached to a size of about 800 mm³ at 30 days after injection. However, knockdown of DDX5 resulted in the suppression of tumor growth (Figure 5c). Thus, DDX5 expression is required for the growth of T cell acute lymphoblastic leukemia in vivo.

Figure 5. DDX5 knockdown reduces growth of human T-ALL leukemia xenograft in nude mice.

(a) DDX5 knockdown in luciferase-expressing HPB-ALL cells (HPB-ALL-luc). DDX5 expression in HPB-ALL-Luc transduced with shRNA against DDX5 (KD) as compared to cells transduced with a nontargeting shRNA (Ctl) was analyzed by Western Blot analysis with anti-DDX5 antibodies. (b) A representative bioluminescent image shows that two mice implanted with HPB-ALL-luc cells transduced with control shRNA and DDX5 shRNA, respectively. The image was taken on Day 24 after leukemic cell injection. (c) Tumor growth was inhibited significantly in mice implanted with DDX5 knockdown cells compared to control (n=5, p<0.01). Tumor volumes were measured every two days starting at day 10 after leukemic cell injection.

DDX5 is highly expressed in human T-ALL patient samples

Since the expression level of DDX5 in T-ALL has not been investigated, we performed data mining and analyzed DDX5 expression using several published datasets from the Oncomine and GEO databases. First, we found that DDX5 transcript levels are significantly up-regulated in T-ALL samples than normal peripheral blood mononuclear cells (PBMC) based on microarray gene expression data of human T-ALL (n=174) and normal PBMC samples (n=74) from the Oncomine Database²² (Figure 6a). Second, DDX5 transcript levels were found to be significantly higher in bone marrow samples from pediatric patients with T-ALL (n=117) in comparison to healthy individuals (n=7) based on the analysis of the GEO GSE26713 dataset²³. Third, DDX5 transcript levels were significantly increased in T-ALL samples (n=57) in comparison with sorted populations of normal human T cells representing various stages of T cell development based on GSE33469 and GSE33470 datasets²⁴ (Figure 6c). These microarray data suggest that DDX5 is highly expressed in human T-ALL patient samples than normal T cells.

Figure 6. DDX5 gene expression is highly expressed in human T-ALL patient samples.

(a) DDX5 transcript levels are significantly higher in T cell acute lymphoblastic leukemia in comparison to peripheral blood mononuclear cells (PBMC) based on microarray gene expression data of T-ALL cases (n=174) and normal PBMC cells (n=74) from the Oncomine database. ***p<0.001. (b) DDX5 transcript levels are significantly up-regulated in bone marrow samples from pediatric patients with T-ALL (n=117) in comparison to healthy individuals (n=7) in GSE26713. **p<0.01. (c) DDX5 transcript levels are significant higher in T-ALL samples (n=57) than in various sorted populations of human thymocytes (n=3). ***p<0.001. The microarray data from the datasets GSE33469 and GSE33470 were normalized using the quantile approach before gene expression analysis. (d) DDX5 is over-expressed in bone marrow aspirates from the patients of T-ALLs by IHC analysis using anti-DDX5 antibodies.

To further determine the relevance of DDX5 in human T-ALLs, we assessed DDX5 protein expression in bone marrow aspirates from three normal individuals and three patients with T-ALLs by immunohistochemical staining. Bone marrow aspirates were formalin-fixed and paraffin-embedded, and then the sections were stained with rabbit anti-DDX5 antibodies. In normal bone marrow aspirates, certain subsets of cells showed positive nuclear DDX5 staining, likely including lymphocytes, granulocytes, and megakaryotes (Figure 6d). In T-ALL samples, strong nuclear DDX5 staining was observed in the majority of T-ALL cells in all three T-ALL specimens (Figure 6d and not shown). In addition, the cellular morphologies seem to be abnormal for T-ALL. Combined with the microarray and IHC analyses, our data strongly support that DDX5 transcript and protein levels are up-regulated in T-ALL samples.

Discussion

Notch signaling is a major oncogenic pathway in the pathogenesis of T-ALLs and is a promising therapeutic target. Therefore, the identification of regulators that modulate Notch signaling is of great importance. In this study, we identified DDX5 as a novel interacting protein for the Notch transcriptional co-activator MAML1. DDX5 interacts with the Notch transcription activation complex via the association with MAML1, and is required for optimal Notch activities in leukemic cells. DDX5 is essential for leukemic cell growth and survival in cultured cells and xenograft tumor models. Importantly, DDX5 is highly expressed in human T-ALL samples. Therefore, DDX5 regulates oncogenic Notch signaling at the transcriptional level and is likely to have a critical role in NOTCH1-mediated T-ALL pathogenesis.

DDX5 is a member of the DEAD box family of RNA helicases^13,14 with diverse cellular functions including RNA processing, mRNA translation, ribosome assembly, and gene transcription.^15,16,25 It was shown that DDX5 co-activates several transcription factors including p53,²⁶ androgen receptor,²⁷ estrogen receptor,²⁸ MyoD,²⁹ Runx2,³⁰ and NFAT5³¹ by mediating chromatin remodeling and transcription initiation complex on target promoters.²⁹ DDX5 is significantly up-regulated in multiple tumors such as breast cancers, prostate cancers, bladder cancers, colon cancers, colorectal tumors, and myeloma,^16,32 which suggest important functions of DDX5 in tumor development and progression. Moreover, DDX5 is responsible for the anti-apoptotic functions of tumor cells when treated with anticancer drugs,^33,34 as depletion of DDX5 in tumor cells abolishes their ability to escape from death and results in acceleration of cell killing by drugs.³⁴ Therefore, DDX5 represents a potential important therapeutic target for cancer treatment.

The expression and function of DDX5 in leukemia had not been investigated previously. In this study, we found that DDX5 is expressed in multiple T-ALL leukemic cell lines (Figure 2a). Analyses of published microarray datasets indicated that DDX5 expression is up-regulated in the bone marrow or peripheral blood specimens derived from human T-ALL patients in comparison to the normal control (Figure 6). Importantly, human T-ALLs expressed significantly enhanced levels of DDX5 than various populations of immature and mature T cells (Figure 6). Therefore, DDX5 might have a regulatory role in the oncogenic events in T-ALL. Indeed, our data showed that DDX5 is required for cell growth and survival of T-ALL cells with constitutive activation of Notch signaling, as DDX5 knockdown inhibited leukemic cell growth and promoted cell apoptosis. Furthermore, we showed that DDX5 depletion results in reduced leukemic xenograft growth. Therefore, DDX5 is an important cellular factor required for leukemic cell growth and survival.

The functional effects of DDX5 knockdown on leukemic cell proliferation and survival are at least partially mediated by the ability of DDX5 to modulate Notch signaling. We showed that DDX5 is recruited to the Notch transcription activation complex, and is physically localized to the Notch responsive promoter, the HES1 promoter. DDX5 regulates Notch-induced transcription, as DDX5 promotes Notch-mediated transcription and DDX5 knockdown reduces expression of Notch signature genes in leukemic cells. However, the exact mechanism by which DDX5 regulates transcriptional activation is still unclear. Previous data indicate that reducing the levels of DDX5 and its homologous protein DDX17 (p72) protein impaired recruitment of the TATA binding protein TBP, RNA polymerase II, and the catalytic subunit of the ATPase SWI/SNF complex, Brg-1.²⁹ Therefore, it is likely that DDX5 recruitment to the Notch target promoter promotes the formation or function of the pre-initiation complex. This might be one of the mechanisms that contribute to MAML1-co-activator function in enhancing Notch transcription.

Intriguingly, inhibition of Notch signaling via targeting different steps of the Notch pathway appears to have distinct effects on cell growth and survival. Notch signaling inhibition via gamma secretase inhibitors (GSIs) that block Notch receptor processing significantly reduces transcription of Notch target genes within 24 hours, but usually has mild and progressive effects on the growth and survival of Notch-mutated T-ALL cell lines.⁵ Consistent with the published data, our data showed that the GSI Compound E inhibited expression of Notch target genes in activating Notch1 mutation-containing KOPT-K1 cells, but had little effects on cell proliferation at day 2 and day 4 under treatment (Supplementary Figure 6). On the other hand, inhibition of Notch signaling via siRNA-mediated NOTCH1 knockdown³⁵ or via blocking MAML1 expression by shRNA (unpublished) or expression of dnMAML1^17,36 results in better and more rapid suppression of cell growth and induction of apoptosis. DDX5 is associated with Notch/MAML1/CSL transcriptional activation complex, which might explain our data that DDX5 depletion functionally resembled MAML1 inhibition, in comparison to GSI-mediated Notch signaling blockade. These data suggest differential mechanisms of action accounting for targeting Notch signaling at various steps of the pathway or Notch-independent roles of MAML1 in the regulation of cell growth and survival. Importantly, it should be noted that DDX5 might regulate leukemic cell growth and survival through other uncharacterized Notch-independent mechanisms due to its ability to interact with other cellular factors. However, it appears that DDX5 has a crucial role in regulating NOTCH1-activated T-ALLs, as DDX5 depletion affected growth and survival of NOTCH1-activated T-ALLs, but not of T-ALL cells without abnormally activated NOTCH1 signaling.

Collectively, we identified DDX5 as a novel MAML1-asociated protein that is required for MAML1 transcriptional co-activation function. DDX5 modulates Notch signaling and is crucial for the growth and survival of T-ALL cells with constitutive activation of Notch signaling, supporting the role of DDX5 as a novel regulator of oncogenic Notch signaling in leukemic cells. Therefore, DDX5 might be critical for Notch-mediated T-ALL pathogenesis and has the potential as a drug target in leukemia. Given the importance of DDX5 in the oncogenesis of multiple cancer types and its function in drugs resistance in tumor cells, development of specific DDX5 inhibitors might hold great promises for the therapy of DDX5 over-expressing tumors including the Notch mutated T-ALL.

Materials and methods

Plasmids

Expression plasmids expressing FLAG-tagged full-length and truncation forms of MAML1, NOTCH1 (ICN1) as well as Notch responsive promoter construct were described previously.^9,11 FLAG-tagged DDX5, Myc-tagged DDX5 and Myc-tagged DDX5 K144R cloned in the pSG5 vectors were kindly provided by Dr. Vittorio Sartorelli.²⁹ DDX5 and DDX5 K144R cDNAs were sub-cloned into a modified pBIND vector to express Gal4 DB-DDX5 fusions.

Antibodies

The following antibodies were purchased from commercial sources: mouse anti-Flag antibody (clone M2, Sigma, St. Louis, MO, USA, 1:500 for immunoprecipitation (IP), 1:1000 for Western Blotting (WB)); anti-Myc (clone 9E10, Santa-Cruz Biotechnology, Santa Cruz, California, USA, 1:1000 for WB); anti-DDX5 (4387, Cell Signaling, Danvers, MA, USA, 1:5000 for WB); anti-DDX5 (PAb204, Upstate, Billerica, MA, USA, 1:100 for ChIP); anti-DDX5 (IHC-00156, Bethyl, Montgomery, TX, USA, 1:200 for IHC); anti-NOTCH1 (SC-6014R, Santa-Cruz, 1:500 for IP, 1:1000 for WB), anti-cleaved NOTCH1 Val 1744 (4147, Cell Signaling, 1:500 for WB), anti-CSL (T-6709, Institute of Immunology Co., Tokyo, JAPAN, 1:500 for WB), anti-Cleaved caspase3 (9664S, Cell Signaling, 1:1000 for WB), anti-MAML1 (A300-673A, Bethyl, 1:5000 for WB), anti-MAML1 (4608, Cell Signaling, 1:500 for IP) and anti-β-actin (A5316, Sigma, 1: 10000 for WB).

Cell culture, transient transfection and reporter assays

Human U2OS Osteosarcoma cells and cervical cancer HeLa cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) plus 10% fetal bovine serum (FBS) and antibiotics. Leukemic cells including KOPT-K1 and HPB-ALL were cultured in RPMI 1640 medium plus 10% FBS and antibiotics. Transfection was performed using Superfect transfection reagent (QIAGEN, Germantown, MD, USA) according to the manufacturer’s instructions. The luciferase-based reporter assays were performed using a dual luciferase kit from Promega (Madison, WI, USA) as previously described.¹¹

Immunoprecipitation and immunoblot analysis

Whole-cell protein extracts were prepared as previously described.⁹ For digestion of nuclear acids, KOPT-K1 protein lysates were treated with DNase (10U/ml) and RNase (10ug/ml) at room temperature for 20 minutes. Immunoprecipitations were performed with antibodies and protein A/G agarose overnight at 4°C. For immunoblot analysis, immunoprecipitates were washed five times and fractionated by SDS-polyacrylamide (PAGE) gels and electrotransferred to NC membranes. The membranes were blocked for 1 h in a buffer containing 10 mM Tris, pH 8.0, 150 mM NaCl, 0.05% Tween 20, and 5% nonfat dry milk. The membranes were incubated with the antibodies overnight at 4°C, then washed and incubated with a horseradish peroxidase-conjugated secondary antibody for 1 h at RT. The protein bands were visualized by enhanced chemiluminescence (Pierce).

Protein complex purification and identification

About 30 mg of whole cell lysates from KOPT-K1 were used for immunoprecipitation with anti-MAML1 antibody and protein A/G agarose at 4°C over night. Protein complexes were washed and eluted from agarose using an elution solution containing MAML1 peptide. The eluted protein samples were separated on the 4–12% NuPAGE Bis-tris gels, and stained with a colloidal blue stain kit (Invitrogen). The MAML1 associated complex was then analyzed by LC-MS/MS mass spectrometry.

GST pull-down assay

The GST fusion proteins GST-MAML1 full-length and 1-302 aa were expressed in BL21 Escherichia coli and purified on glutathione-sepharose beads according to the manufacturer’s protocol. Equivalent amounts of purified GST fusion proteins or GST bound to glutathione-sepharose beads were incubated with 293T cell lysate expressing transfected FLAG-tagged DDX5 proteins overnight at 4°C. The beads were washed extensively and the proteins bound to the beads were resolved on SDS-polyacrylamide (PAGE) gels and detected by Western blot analysis.

Lentiviral-mediated shRNA knockdown

The pLKO.1_DDX5-shRNA and the control luciferase shRNA constructs were purchased from OpenBiosystems. Lentiviruses were produced as previously described.³⁷ In brief, 293FT cells were transfected with shRNA targeting DDX5 or luciferase control plasmid together with packing plasmid psPAX2 and envelope plasmid pMD2.G. Viral supernatants were harvested at 48 h and 72 h post-transfection, target cells were infected in the presence of 4 ug/ml polybrene for 8 h twice for two consecutive days and then selected with puromycin (0.2 μg/ml, Sigma).

Real-time RT-PCR

Real-time RT-PCR was performed as previously described.³⁷ RNA was isolated and then reverse transcribed into cDNA using a GeneAmp RNA PCR kit (Applied Biosystems). Real-time PCR was performed using the StepOne Real-Time PCR System (Applied Biosystems) with the SYBR Green PCR Core Reagents Kit (Applied Biosystems, Carlsbad, CA, USA). GAPDH was used as an internal control to normalize gene expression levels. The following primers were used: HES1 primers (Forward, 5′ TCAACACGACACCGGATAAA 3′; Reverse, 5′ TCAGCTGGCTCAGACTTTCA 3′); DTX1 primers (Forward, 5′ CTTCCCTGATACCCAGACCA 3′; Reverse, 5′ TCCTCTTGCGGTGAACTTCT 3′); HEY1 primers (Forward, 5′ tcggctctaggttccatgtc 3′; Reverse, 5′ ctgggtaccagccttctcag 3′); MYC primers (Forward, 5′ ggactatcctgctgccaaga 3′; Reverse, 5′ cgcctcttgacattctcctc 3′); DDX5 primers (Forward, 5′ GCCGGGACCGAGGGTTTGGT 3′; Reverse, 5′ CTTGTGCTGTGCGCCTAGCCA 3′); and GAPDH primers (Forward, 5′ CAATGACCCCTTCATTGACC 3′; Reverse, 5′ GACAAGCTTCCCGTTCTCAG 3′).

Proliferation assays

Cell growth was determined by counting cell numbers in culture. Briefly, 2 ml of 1×10⁵/ml cells were cultured in RPMI1640 containing 10% FBS in the 6-well plates at day 0, and the cell numbers were scored by trypan blue exclusion at day 2 and day 4.

Cell cycle analysis

Cell cycle profiles were analyzed by BrdU/7-AAD staining using the BD Pharmingen™ BrdU Flow Kit (BD Biosciences) according to the manufacturer’s instructions. After staining, a minimum of 10000 events were collected to create each DNA content histogram with Accuri C6 Flow Cytometer (BD Biosciences).

Apoptosis analysis

Cell apoptosis were detected by measuring Caspase 3/7 activity with the Caspase-Glo 3/7 Assay kit (Promega) and the Annexin V/PI FACS analysis (BD Biosciences) according to the manufacturer’s instructions. For the Caspase 3/7 activity analysis, ~1×10⁴ cells (20 μl) and an equal volume of Caspase-Glo 3/7 reagent were added to a well in the 96-well plate. After incubation for 1–2 h in room temperature, the luminescence of each sample and blank control were measured in a plate-reading luminometer (FLUOstar OPTIMA). For Annexin V/PI FACS analysis, the cells were stained with Annexin V-FITC and Propidium Iodide for 15 min at room temperature in the dark and then analyzed with Accuri C6 Flow Cytometer (BD Biosciences).

Chromatin Immunoprecipitation (ChIP)

The ChIP assays were performed based on the Millipore (Billerica, MA, USA) ChIP protocol with minor modifications as previously described.³⁷ Briefly, cells were fixed with 1% formaldehyde for 10 min at room temperature and sonicated to shear DNA to about 200 and 800 bp. The DNA protein complex was then immunoprecipitated with the DDX5 antibodies or control IgG. The ChIP DNA was purified and eluted with 100ul of H₂O. ChIP DNA (2.5ul) was used for the real-time PCR analysis using the primers flanking the CSL-binding site of the Notch target HES1 promoter. The primer sequences are: Forward, 5′ cgtgtctcctcctcccatt 3′; and Reverse, 5′ ggcctctatatatatctgggactgc 3′.

Immunohistochemical staining

Formalin-fixed paraffin-embedded bone marrow from The University of Texas MD Anderson Cancer Center with an approved IRB. These slides were heated at 50°C for 1 h before being deparaffinized and rehydrated. Endogenous peroxidase activity was blocked with 3% hydrogen peroxide for 10 min. Optimal staining with rabbit anti-DDX5 (IHC-00156, Bethyl) required 25 min. of heat antigen retrieval in 10mM Citrate buffer pH 6.0. Sections were sequentially blocked with normal serum, avidin and biotin prior to the application of rabbit anti-DDX5 (1:200) overnight at 4°C. Slides were stained using the ABC-Elite kit (Vector Labs, Burlingame, CA, USA) and counter-stained with Hematoxylin (SurgiPath, Buffalo Grove, IL, USA).

In vivo tumor xenograft

Tumor xenograft studies were performed following a protocol approved by IACUC (Institutional Animal Care & Use Committee) of University of Florida. HPB-ALL cells that had been engineered to stably express firefly luciferase (HPB-ALL-Luc) by transducing with pMSCV-GFP-Luc retroviruses were infected with lentiviruses carrying control shRNA, or DDX5-targeting shRNA. A total of 5×10⁶ leukemia cells were diluted in 100 μl 50% Matrigel (BD Biosciences)/50% PBS and injected subcutaneously into the right flank of 6-week-old female NCR nude mice. For bioluminescence imaging, mice were given 150 μg/g of D-luciferin in PBS by i.p. injection. Fifteen minutes after injection, bioluminescence was imaged with a Xenogen In-vivo Imaging System (Caliper Life Sciences, Hopkinton, MA, USA). Tumor volumes were measured in two dimensions (length and width) using Dial Caliper and calculated using the formula: tumor volume = (length × width²) × 0.5. After 10 days post-implantation, tumor volumes were measured every two days and bioluminescence images were taken once a week.