Identification of C/EBPβ Target Genes in ALK+ Anaplastic Large Cell Lymphoma (ALCL) by Gene Expression Profiling and Chromatin Immunoprecipitation

C/EBPβ (CCAAT enhancer binding protein) is a transcription factor that plays a crucial role in survival and transformation of ALK+ anaplastic large cell lymphoma (ALCL). The aim of this study was to identify the downstream targets of C/EBPβ responsible for ALK-mediated oncogenesis. C/EBPβ was knocked down in ALK+ ALCL cell lines with a C/EBPβ-shRNA, followed by gene expression profiling (GEP). GEP analysis revealed a reproducible signature of genes that were significantly regulated by C/EBPβ. Classification into biological categories revealed overrepresentation of genes involved in the immune response, apoptosis and cell proliferation. Transcriptional regulation by C/EBPβ was found in 6 of 11 (BCL2A1, G0S2, TRIB1, S100A9, DDX21 and DDIT4) genes investigated by chromatin immunoprecipitation. We demonstrated that BCL2A1, G0S2 and DDX21 play a crucial role in survival and proliferation of ALK+ ALCL cells. DDX21, a gene involved in rRNA biogenesis, was found differentially overexpressed in primary ALK+ ALCL cases. All three candidate genes were validated in primary ALCL cases by either immunohistochemistry or RT-qPCR. In conclusion, we identified and validated several key C/EBPβ-regulated genes with major impact on survival and cell growth in ALK+ ALCL, supporting the central role of C/EBPβ in ALK-mediated oncogenesis.


Introduction
The transcription factor CCAAT/enhancer binding protein beta (C/EBPb) is one of six members of the C/EBP leucine zipper transcription factor family. C/EBPb is an intronless gene, which is transcribed as a single mRNA that can produce at least three isoforms (liver-enriched activating proteins LAP (46 kDa) and LAP* (48 kDa) and liver-enriched inhibitory protein LIP (21 kDa)). It is involved in a number of cellular processes, including differentiation, proliferation, inflammatory responses and metabolism [1,2]. Furthermore, C/EBPb has been implicated in oncogene-mediated tumorigenesis and apoptosis resistance in solid tumors [3,4].
Recently, we reported that C/EBPb is overexpressed in anaplastic lymphoma kinase (ALK)+anaplastic large cell lymphoma (ALCL), and demonstrated that its expression is dependent on ALK kinase activity [5,6]. ALK+ ALCL is a distinct subtype of non-Hodgkin's lymphoma with unique morphologic and immunophenotypical features. ALK+ ALCL is characterized by the t(2;5) chromosomal translocation, which juxtaposes the nucleophosmin (NPM) gene and the ALK gene, resulting in the expression and constitutive activation of ALK protein [7,8]. Subsequent studies have shown that about 20% of ALK+ ALCLs contain variant translocations in which the ALK gene is fused to other partner genes [7,9]. ALK-fusion proteins interact with many adaptor proteins and activate several key signaling pathways involved in cell proliferation, transformation and survival, including STAT3, AKT/mTOR and ERK1/2 signaling pathways [8,10,11]. Furthermore it was shown that NPM-ALK exerts HuR-mediated posttranscriptional control on C/EBPb gene expression that leads to increased C/EBPb mRNA stability and translation in ALK+ ALCL [12].
In our previous study, we showed that the expression of C/ EBPb in ALK+ ALCL is controlled primarily by the STAT3 pathway, whereas its phosphorylation and activation is partially dependent on the MAPK pathway. Furthermore, we demonstrated a critical role of C/EBPb in the proliferation and survival of ALK+ ALCL cells [6]. Because C/EBPb seems to be central to ALK transformation, the aim of the current study was to identify downstream targets of C/EBPb to gain insight in the pathogenesis of ALK+ ALCL. We now demonstrate using gene expression profiling (GEP) and chromatin immunoprecipitation (ChIP) analyses that C/EBPb regulates important genes responsible for cell proliferation and survival in ALK+ ALCL. We explore in some detail, one of these genes, the RNA helicase DEAD box polypeptide 21 (DDX21), a member of the DExD/H box family of proteins that play important roles in RNA metabolism.

Ethics Statement
Ethics approval for the study (620/2011BO2) was obtained from the Ethics Committee at the Medical Faculty, University Tübingen. Written informed consent was obtained for the lymphoma specimens used in this study.

Plasmids, Cell Culture and Patient Samples
Oligonucleotides containing the C/EBPb-shRNA [6] and DDX21-shRNA (59-TGATAAGACTGAAGAGATA-39) sequences were cloned into pSuper (Oligoengine, Seattle, WA, USA) and pFUGW as previously described [6,13]. pGIPZ lentiviral vectors containing a BCL2A1-and G0S2-shRNA were purchased from Open Biosystems (Open Biosystems Products/ Thermo Fisher Scientific, Huntsville, AL, USA). The In-Fusion TM Advantage PCR Cloning Kit (Clontech Laboratories, Mountain View, CA, USA) was used to clone the template for the expression of the C/EBPb isoforms LAP* and LAP into the vector pRRL.PPT.SF.i2GFPp [14] (for details see Methods in File S1). The ALK+ ALCL cell lines (SUDHL-1, KiJK and Karpas 299) were provided by Mark Raffeld (National Cancer Institute, NIH, Bethesda, MD, USA), and cultured as recently described [6]. SUDHL-1 and Karpas 299 were purchased from the American Type Culture Collection (ATCC) and KiJK was obtained from the author [15]. All three cell lines have been authenticated and are suitable for in vitro model system for ALCL [16]. The ALK-ALCL cell lines Mac-1 was provided by Eva Geißinger (University of Würzburg, Germany), and Mac-2A by Olaf Merkel (Deutsches Krebsforschungszentrum (DKFZ), Heidelberg, Germany). The mantle cell lymphoma (MCL) cell line Granta 519 was obtained from the cell bank of the mantle cell lymphoma research initiative and consortium after authentication and now is available from ATCC. Eight ALK-and ten ALK+ ALCL primary samples were collected from the files of the Institute of Pathology (University Hospital Tübingen, Germany). All cases were comprehensively immunophenotyped, as part of the diagnostic work-up, and were classified following the recommendations of the World Health Organization classification for tumors of haematopoietic and lymphoid tissues [9].

Virus Production and Viral Infections
Production of virus containing pFUGW was performed as recently described [17][18][19]. Production of virus containing the pGIPZ vectors was performed transfecting 36610 6 cells in 10 cm plates using 30 ml TransIT 2020, 2 mg pHCMV-G, 6 mg pCMVdeltaR8.9 and 8 mg of the respective pGIPZ derivate according to the manufacturers manual (Mirus Bio, Madison, MI, USA). Production of virus containing the pRRL.PPT.SF.i2GFPp vectors was also performed (see Methods in File S1). Infection of lymphoma cells was performed as recently described with modified centrifugation conditions (800 g, 32uC, 90 min) [17]. To achieve efficient knockdown of DDX21, infection was repeated after 24 h (MOI of 90). Transduction efficiency was determined as previously specified [6,17].

Cell Proliferation and Viability Assay (MTS Assay) and Analysis of Apoptosis
Cell viability and growth retardation was determined by the MTS cell proliferation assay (AQueous CellTiter 96, Promega, Mannheim, Germany) [6]. Cell cycle analysis by flow cytometry was performed according to Nuesse et al. (FACSCalibur, BD, Franklin Lakes, NJ, USA) [20]. Analysis of apoptosis was done by annexin V (AnnexinV-APC, Invitrogen) and propidium iodide (PI) (Sigma-Aldrich, St. Louis, MO, USA) stainings according to standard protocols, followed by flow cytometry.

RNA Isolation and GEP
Total RNA was isolated using the RNeasy Mini kit (Qiagen, Hilden, Germany) including DNase treatment in cell lines and phenol-chloroform extraction for formalin-fixed paraffin embedded tissues in primary cases [21]. For GEP RNA quality was examined using the Agilent 2100 Bioanalyzer and only high quality RNA (RIN.8) was used for subsequent analyses.
Three biological replicates (transduction with control shRNA or shRNA directed against C/EBPb) were used for the analysis. Total RNA (1 mg) was isolated 72 hours after transduction and amplified using a single round amplification kit (Affymetrix, Santa Clara, CA, USA). 10 mg of amplified aRNA were hybridized on Affymetrix U133 plus 2.0 arrays containing about 54.000 probe sets. Staining and scanning was done according to the Affymetrix standard protocol. The Bioconductor software package implemented in CARMAweb [22] was employed to calculate probe set summaries (RMA) and for normalization using the quantile algorithm. For statistical analysis the moderated limma test was applied in combination with the Benjamini-Hochberg correction for multiple testing. Probe Sets with a False Discovery Rate (D) C/EBPb signature of SUDHL-1 and KiJK cells derived by C/EBPb-shRNA. Heatmaps show the expression pattern of the 114 genes which were significantly (FDR,10%) regulated after C/EBPb knockdown in both cell lines, SUDHL-1 and KiJK, depicting transduced cells with C/EBPb-shRNA and controls in triplicates. The 23 candidate genes whose regulation was validated by RT-qPCR are designated. The coloured log2 scale bar represents relative gene expression changes. (E) GO term enrichment analysis was used to assess the biological functions of the 114 differentially expressed genes [23]. Selected terms (p,0.01) and the number of genes associated with them are shown. SUDHL-1 = uninfected cells, pF = empty virus, pF-C/ EBPb = virus containing the C/EBPb shRNA sequence. doi:10.1371/journal.pone.0064544.g001 (FDR),10% were considered as significant. Heatmaps and dendrograms were generated with tools provided by CARMAweb.
Gene Ontology (GO) terms [23] and pathway analysis were done with the Genomatix Pathway System (Genomatix, Munich, Germany) and Ingenuity IPA software was used to create one of the figures. Microarray data has been submitted to GEO (GSE23368): http://www.ncbi.nlm.nih.gov/geo/query/acc. cgi?token = hxoppwqssooewvo&acc = GSE23368. Real-time Quantitative RT-PCR (RT-qPCR) C/EBPb mRNA levels were quantified using TBP as housekeeping gene as previously specified [6,24]. For all other genes, primers and probes were combined applying the Universal ProbeLibrary System (Roche Applied Science, Penzberg, Germany) (Table S1 in File S1). 500 ng RNA were transcribed into cDNA using Superscript II reverse transcriptase (Invitrogen, Carlsbad, CA, USA) and a mix of Oligo(dT) primer (Promega, Madison, WI, USA) and random hexamers (Roche Applied Science, Penzberg, Germany) in a final volume of 50 ml according to the manufacturer's instructions. RT-qPCRs were carried out in duplicates with the LightCycler 480 Probes Master using a LightCycler 480 System for detection (Roche Applied Science). Data were analyzed using the 2 2DDCp method [5].
Chromatin Immunoprecipitation (ChIP) Assay and Luciferase Reporter Assay SUDHL-1 cells were fixed, nuclei were isolated and chromatin was sheared by sonication using the ChIP-IT Express kit according to the manufacturer's instructions (Active Motif, Carlsbad, CA, USA). Sheared chromatin was immunoprecipitated using C/EBPb antibody (3 mg, Santa Cruz Biotechnology) and IgG (3 mg, Sigma-Aldrich). Isolated DNA was analyzed by qPCR (LightCycler 480 (Roche Applied Science)), and amplification of 50 sequences (primers Table S2 in File S1) of 11 genes (BCL2A1, TRIB1, G0S2, S100A9, DDIT4, TM4SF1, CDKN2A, PTPRC, DDX21, JUN, CCL20) with potential C/EBPb binding sites (Bibliosphere, Genomatix, TFSEARCH (http://www.cbrc.jp/ research/db/TFSEARCH.html)) was performed using the Quan-tiTect Sybr Green PCR Kit (Qiagen) according to the manufacturer's protocol. Some qPCRs covered more than one regulatory site. To normalize for variations in the amount of starting DNA, a DCp value was calculated (DCp(input)-DCp(ChIP)). DNA quantities were expressed as percentages of corresponding input (2 2DDCp x100). Mean values of three biological replicates were calculated. Statistical analysis was carried out using t-test. To get additional evidence for C/EBPb directly targeting the regulatory site of the intron region of DDX21, luciferase reporter assays were performed (see Methods in File S1).

Metabolic Labeling of Nascent Pre-rRNA with 32P
SUDHL-1 cells were metabolically labeled with 32 P after transduction with DDX21-shRNA. 1.5610 6 cells were cultured with phosphate-free RPMI 1640 (MP Biomedicals, Singapore) for 3 h. Medium was replaced with phosphate-free medium containing 40 mCi/ml [ 32 P]Orthophosphate (PerkinElmer, Waltham, MA, USA) and cells were incubated for an additional 1.5 h. The chase experiment was done by cultivating the labeled cells with nonradioactive standard RPMI 1640 supplemented with 10% FCS at 0, 2, 3.5 and 5 h. Total RNA was isolated using the RNeasy Mini kit and 250 ng of total RNA was resolved on a formaldehyde-containing 1% agarose gel. The gel was dried and radioactive bands were detected by autoradiography.

Efficient Transfer of C/EBPb-shRNA into ALCL Cell Lines Using a Lentiviral Vector
Two ALK+ ALCL cell lines with strong C/EBPb expression -SUDHL-1 and KiJK -were transduced with C/EBPb-shRNA. Flow cytometric analysis showed that C/EBPb-shRNA was effectively transduced into the cell lines with a high infection rate of almost 100%. Three days after transduction, C/EBPb was successfully downregulated, as demonstrated by Western blot and RT-qPCR ( Figure 1A-B and Figure S1A-B). C/EBPb mRNA was reduced by 73% compared to untreated control cells. To further evaluate the effect of C/EBPb downregulation, growth curves were generated from treated cells up to six days after infection. A comparison between control and C/EBPb-shRNA-infected cells on day six of infection revealed growth retardation of 81.3% in SUDHL-1 ( Figure 1C), and of 72.4% in KiJK ( Figure S1C).

Identification of Differentially Expressed Genes After C/ EBPb Downregulation
In order to identify the downstream targets of C/EBPb in ALK+ ALCL that might be responsible for survival and proliferation, GEP was performed after C/EBPb silencing in the two ALK+ ALCL cell lines SUDHL-1 and KiJK, and compared with mock-treated cells. Statistical analysis of the gene expression data resulted in 595 genes (corresponding to 909 probe sets) that were significantly (FDR,10%) regulated in SUDHL-1 after C/ EBPb knockdown (314 upregulated, 278 downregulated, 3 ambigous). In KiJK cells, 392 significantly regulated genes (262 upregulated, 130 downregulated; corresponding to 582 probe sets) were identified. Among the significantly regulated genes 129 were regulated more than 2-fold in SUDHL-1 and 61 in KiJK. To define a consistent signature of C/EBPb-dependent transcripts, we focused on genes that were significantly regulated after C/EBPb silencing in both cell lines. The majority of these 114 genes (corresponding to 166 probe sets, Table S3 in File S1) showed similar regulation (up or down) in SUDHL-1 and KiJK ( Figure 1D). To address the biological functions of the 114 differentially expressed genes, we analyzed their associated GO terms to obtain significantly (p,0.01) enriched terms (Table S4 in File S1) [23]. This analysis revealed the overrepresentation of terms related to immune response, apoptosis and cell proliferation ( Figure 1E).
To identify the most promising candidates of C/EBPbmediated oncogenesis, 26 genes that were strongly regulated (14 genes) and/or had potential C/EBPb promoter binding sites (17 genes) (Bibliosphere, TFSEARCH (http://www.cbrc.jp/research/ db/TFSEARCH.html)) were selected for further analysis. Validation of the candidate genes by RT-qPCR confirmed the regulation of 23 of the 26 genes in both cell lines ( Table 1). Some of these genes showed remarkable pathway connections, as revealed by the ''Ingenuity pathway'' program. The three best candidate genes -G0S2, DDX21 and BCL2A1showed very high mRNA expression levels (Figure 2A), and followed the same pattern as C/EBPb expression in the three analyzed ALK+ ALCL cell lines. SUDHL1 showed the highest mRNA expression levels of all three genes, whereas Karpas 299 showed the lowest levels of expression. Western blot analysis corroborated the expression of BCL2A1, DDX21 and G0S2 in all three analyzed ALK+ ALCL cell lines. Additionally, TRIB1 was expressed in all three cell lines but at lower levels ( Figure 2B). To further validate the C/EBPb dependency of these genes, Western blot analysis was performed after C/EBPb silencing. Coordinate with the reduction of C/ EBPb protein, there was a clear decrease in the protein levels of BCL2A1, DDX21, G0S2 and TRIB1 ( Figure 2C) with the strongest correlation observed for DDX21. This suggested that these four genes were regulated by C/EBPb.

Direct Gene Targets of C/EBPb
We next investigated 50 potentially transcriptional regulatory locations (Table S2 in File S1) of eleven genes, which had been selected using the web tool TFSEARCH to predict proximal promoter elements and Bibliosphere to predict regulatory elements that can be distributed further over the respective gene locus, to identify C/EBPb binding sites by qPCR after ChIP analysis in SUDHL-1 cells. C/EBPb binding was detected on at least one site of the regulatory regions of BCL2A1, TRIB1, G0S2, S100A9 and DDIT4 consistent with direct regulation by C/EBPb. Representative qPCR results of the different regulatory sites of each gene are shown in Figure 2D. C/EBPb binding was detected in more than one location in several genes. Positive signals were   in three of four different regulatory regions of the BCL2A1 gene, two of three locations within the G0S2 gene and two of seven locations within the S100A9 gene, whereas C/EBPb binding was detected in a single regulatory region of TRIB1 and DDIT4. Surprisingly, no signals were detected analyzing the predicted DDX21 sites, although DDX21 was a promising candidate gene for direct regulation by C/EBPb because of its high mRNA and protein expression and strong correlation with C/EBPb expression in ALK+ ALCL cells. However, all six potential binding sites in the proximal region 21585 to 2130 59 of the transcriptional start site of the DDX21 gene investigated were negative ( Figure 3A). Considering that regulatory elements could have escaped the Bibliosphere screening, the TFSEARCH analysis was extended to intron regions of the DDX21 gene. An additional regulatory site in the first intron of the gene was found (Table S2 in File S1) that showed a significant positive signal by ChIP analysis (Figure 3A), providing strong evidence for the direct transcriptional regulation of DDX21 by C/EBPb. To further demonstrate that DDX21 is a direct target of C/EBPb, luciferase reporter assay was performed analyzing the regulatory site located in the first intron of DDX21. Overexpression of C/EBPb (transient expression of the complete coding sequence produces mainly isoform LAP* and LAP) in HEK293T cells led to a significant increase in luciferase activity in comparison to the transfection of each of the plasmids separately ( Figure 3B). To further confirm the C/EBPbdependent expression of DDX21, the isoforms LAP* and LAP, which have an activating transcriptional function, were overexpressed in the ALK-ALCL cell line Mac-1 and the ALK+ ALCL cell line SR786, which does not express C/EBPb. Lentiviral vectors containing the coding sequence of LAP* and LAP were efficiently transduced in the cells, as demonstrated by expression of the GFP reporter gene ( Figure 3C). In both cell lines a strong upregulation of the isoforms LAP* and LAP was detected three days after infection together with clear upregulation of DDX21 ( Figure 3D). Taken together, these data confirm that DDX21 is a direct target of C/EBPb. qPCRs of the potential regulatory sites of the remaining five genes investigated (TM4SF1, CDKN2A, PTPRC, JUN, CCL20) showed no signals, indicating the absence of C/EBPb binding. Consequently, these genes seem to be regulated by C/EBPb by alternative mechanism(s) other than direct transcriptional activation.

BCL2A1, G0S2 and DDX21 Contribute to the Survival and Proliferation of ALK+ ALCL Cells
To evaluate the effect of BCL2A1 in ALK+ ALCL cells, BCL2A1 was silenced with a specific BCL2A1-shRNA construct in SUDHL-1 cells ( Figure 4A). RT-qPCR confirmed the BCL2A1 mRNA downregulation of approximately 60% compared to control cells ( Figure 4B). Growth curves were generated from treated cells with 99% infection rate up to 6 days after infection. Downregulation of BCL2A1 mRNA led to a strong growth retardation of almost 60% after two days of infection when compared to control cells ( Figure 4C). To see whether the downregulation of BCL2A1 induces apoptosis, flow cytometric analysis with annexin V/propidium iodide was performed three days after infection. BCL2A1-shRNA-infected cells revealed increased cell death (34.1%) when compared to the control cells (11.5% and 11.9% respectively) ( Figure 4D). These findings demonstrate the important role of BCL2A1 in the survival and sustained growth of ALK+ ALCL cells. To investigate whether BCL2A1 was also expressed in primary cases, we analyzed 10 ALK+ and 8 ALK-ALCL cases by immunohistochemistry ( Figure 4E). All cases analyzed were similarly positive for BCL2A1, suggesting that the overexpression of BCL2A1 is important for promoting survival in both tumor types.
Since G0S2 has been implicated in cell cycle control, and was one of the genes with the highest mRNA expression, its role in the survival and proliferation of ALK+ ALCL cells was investigated. G0S2 was silenced in SUDHL-1 cells with a specific G0S2-shRNA construct ( Figure 5A). The knockdown of G0S2 resulted in a 72% reduction of mRNA as demonstrated by RT-qPCR ( Figure 5B). Downregulation of G0S2 in SUDHL1 cells resulted in growth retardation of 23% after 5 days of infection when compared to the untreated control cells ( Figure 5C). Only a slight increase in apoptosis was detected after GOS2 knockdown, as demonstrated by annexin V/propidium iodide assay (16.2% vs. 4.4%) ( Figure 5D). Cell cycle analysis demonstrated a decrease in S and G2/M phase of 14.7% and 16.0%, respectively, and an increase in G0/G1 of 19.2% after four days of infection ( Figure 5E). The subG1 fraction showed 27,1% of dead cells in the GOS2 downregulated cells compared to 9,3% in the untreated control cells. These results reflect the mild apoptosis found with the annexinV/propidum iodide assay. Western blot analysis for caspase 3 showed a mild increase in cleaved caspase 3 after GOS2 downregulation ( Figure 5F). Taken together, these results show a role of G0S2 in sustaining proliferation of ALK+ ALCL cells and only a minor role in inducing apoptosis. The expression of G0S2 protein was investigated in primary cases; however, the available antibodies gave a non-specific reaction in paraffin-embedded tissue. Therefore, G0S2 mRNA levels were analyzed in six ALK+ vs. eight ALK-primary cases ( Figure 5G). ALK+ ALCL cases showed a significantly higher G0S2 mRNA expression than ALKcases (p = 0.017).
As DDX21 had been proved to be involved in pre-rRNA processing and participated in cell growth [25][26][27][28], the possible role of DDX21 in cell proliferation and survival as well as in rRNA processing of ALK+ ALCL cells was further investigated. SUDHL-1 cells were transduced with a specific DDX21-shRNA construct ( Figure 6A). Because of the very high DDX21 mRNA expression levels, a second infection was performed after 24 hours, which resulted in a further mRNA reduction of 92% compared to 66% after the first infection ( Figure 6B). Efficient downregulation of DDX21 mRNA and protein showed growth retardation of 65.3% in SUDHL-1 cells after 7 days ( Figure 6C 11.4/11.9% annexin V+/PI+ and 4.2/2.6% annexin V2/PI+) ( Figure 6D). To investigate whether the impact of DDX21 knockdown in ALK+ ALCL cells was unique to these cells, we performed DDX21 knockdown in other cell lines (Mac-2A, ALK-ALCL and the MCL Granta 519). The efficient downregulation of DDX21 mRNA and protein in these two cell lines after the first infection produced also growth retardation; however, the effect was stronger in Mac-2A cells (94.3%) than in in Granta 519 cells (64.1%) ( Figure S2). These results suggest that DDX21 is necessary to maintain proliferation in different lymphoma cells. To analyze the possible implication of DDX21 in rRNA biogenesis and cell metabolism of ALK+ ALCL, we investigated whether processing of pre-rRNA was interrupted by silencing DDX21. Autoradiography of RNA purified from ALK+ ALCL cells with strong DDX21 knockdown demonstrated a clear delay in processing of 32Slabeled pre-rRNA. Downregulation of DDX21 resulted in a 70 and 74% decrease in the levels of 28S and 18S rRNA, respectively after a 3.5 h chase compared with the control cells ( Figure 6E).
To determine whether the high expression of DDX21 was also present in primary cases, ten ALK+ and eight ALK-ALCL cases were analyzed by immunohistochemistry. All ALK+ ALCL cases showed a strong nucleolar expression of DDX21 in the tumor cells ( Figure 6F). In contrast, ALK-ALCL cases showed normal expression of DDX21 comparable to the reactive lymphoid cells. These results clearly demonstrated differential expression of DDX21 in ALK+ ALCL compared to ALK-ALCL primary cases, and confirmed the strong interdependence of DDX21 and C/EBPb expression observed in the cell lines.

Discussion
We recently reported that NPM-ALK induces C/EBPb expression primarily through the STAT3 signalling pathway and that C/EBPb plays a central role in ALK-mediated transformation in ALK+ ALCL [6,29]. In this study, we aimed to identify the downstream targets of C/EBPb that might contribute to ALKmediated lymphoid transformation. To achieve this objective, GEP was performed after C/EBPb silencing in ALK+ ALCL cell lines. GEP produced a reliable and reproducible expression signature of 114 genes, which were significantly regulated by C/ EBPb in two ALK+ ALCL cell lines with strong C/EBPb expression -SUDHL-1 and KiJK. Overrepresentation of genes mainly involved in immune system processes, apoptosis and cell proliferation reflected the central role of C/EBPb in the proliferation and survival of ALK+ ALCL cells. Using ChIP analysis and highly effective shRNA sequences, we investigated the pathogenic role of several C/EBPb-regulated genes and assessed their individual contribution to the maintenance of the malignant phenotype in ALK+ ALCL. We demonstrated that three genes, BCL2A1, G0S2, and DDX21 are necessary for sustaining survival and proliferation of ALK+ ALCL cells, and found several other potentially interesting genes.
GEP is a useful tool not only in defining the prognosis of tumors and for discovering diagnostic genomic classifiers, but also for the identification of regulated genes and signaling pathways. Therefore, we decided to take this approach together with a highly effective C/EBPb-shRNA capable of silencing the expression of C/EBPb in ALK+ ALCL cells to generate a reliable signature of genes regulated by C/EBPb. Of the original 114 genes identified by GEP, 23 of 26 selected candidate genes were confirmed by RT-qPCR to be regulated by C/EBPb. To further validate our data, we compared our C/EBPb-signature with published ALKsignatures. Not surprisingly, only a limited number of overlapping genes among these studies were found (TXNIP, AQP9, ELAC2, UGCG, UBE2H, BCL2A1, DDX21 and CCL20, G0S2, S100A9, UPK1B) [29][30][31]. Three of the tightly regulated C/EBPb candidate genes (BCL2A1, DDX21 and CCL20) were also identified in the ALK-signature [29]. Moreover, a comparative study of ALK+ ALCL to other T-NHLs also had demonstrated that some of our candidate genes (G0S2, S100A9, UPK1B) are differentially expressed in ALK+ ALCL tumors [31]. The overlapping detection of these genes in the different signatures indicates that at least some of them are most probably important mediators of NPM-ALK oncogenesis. Furthermore, our results underscore the relevance and contribution of C/EBPb transcriptional activity to the expression signature of ALK+ ALCL cells.
In order to understand better the role of C/EBPb in ALK+ ALCL pathogenesis, genes regulated by C/EBPb-knockdown were further characterized by a Gene Ontology term enrichment analysis. Significantly overrepresented terms were related to the immune system process (29%), cell proliferation (21%), apoptosis (21%) and inflammatory response (13%), which is in agreement with the well-known functions of C/EBPb in macrophage differentiation, inflammatory response and fat metabolism [2]. Of the 114 candidate genes that composed the C/EBPb signature, 26 genes were selected for further analysis based on its high mRNA expression and/or strong regulation by C/EBPb, possible pathway connections and potential promoter binding sites for C/ EBPb.
To confirm potential direct transcriptional targets of C/EBPb, ChIP analysis was performed in 12 of the 23 validated candidate genes. Promoter binding of C/EBPb was detected in six of the investigated genes, which included BCL2A1, G0S2, TRIB1, S100A9, DDIT4, and DDX21. The antiapoptotic gene BCL2A1 is of particular interest because it is a member of the Bcl-2 protein family largely restricted to the hematopoietic compartment, and specifically a direct target of pre TCR-signalling, thereby, promoting the viability of T-cells. BCL2A1 is a direct transcription target of NF-kappa B (NF-kB) in response to inflammatory stimuli and is upregulated by different extracellular signals such as B-cell receptor stimulation, and CD40 signaling [32]. BCL2A1 protects cells from various death stimuli including TNFa stimulation, p53 activation, T-and B-cell receptor triggering and IL3 deprivation, indicating a cytoprotective function essential for lymphocyte activation, as well as survival [32]. Importantly, BCL2A1 was described together with C/EBPb as critical target of ALK-activity, necessary for sustaining the survival and/or growth of ALK+ ALCL cells [29]. Our results corroborate and extend these previous findings establishing BCL2A1 as a direct transcription target of C/EBPb necessary to promote survival of ALK+ ALCL cells. We confirmed the antiapoptotic key role of BCL2A1 in ALK+ ALCL cells, which showed significant growth retardation (60%) and cell death (34.1%), 48 hours after BCL2A1 silencing. Of note, no difference in BCL2A1 protein expression was found in ALK+ vs. ALK-ALCL primary cases. Accordingly, BCL2A1, in addition to being a transcriptional target of C/EBPb, can also be regulated by other transcription factors such as NF-kB or AP1 [33]. Interestingly, it has been shown that NF-kB activity is either very low or not present in ALK+ ALCL and that its activity is blocked directly by NPM-ALK [34]. In contrast, NF-kB is constitutively expressed in many peripheral T-cell lymphomas and in classical Hodgkin lymphoma [35,36]. NF-kB activation has been related to oncogenesis through several mechanisms, mainly promoting cellular proliferation and inhibition of apoptosis [37]. Therefore, it is tempting to speculate that NF-kB might be one alternative pathway responsible for BCL2A1 induction in ALK-ALCL tumors. Regardless of the mechanism, the universal expression of BCL2A1 implies that this protein is necessary for survival in both ALK+ and ALK-ALCL, and therefore, it might be a potential candidate for targeted therapy.
G0S2 is a gene differentially expressed in primary ALK+ ALCL with very high mRNA and protein expression levels and is tightly regulated by C/EBPb, but whose biological functions are not clearly understood. G0S2 is widely expressed in normal human tissues and is present at particularly high levels in peripheral blood [38]. G0S2 was originally reported to have a role in the re-entry of cells from G0 into the G1 phase of the cell cycle, and therefore, involved in promoting cell proliferation [39]. Accordingly, the inhibition of G0S2 resulted in cell arrest in endometrial cells [40]. However, a recent study suggested that the upregulation of G0S2 was associated with cell-cycle withdrawal rather than cell proliferation [41]. Moreover, ectopic expression of G0S2 was shown to induce apoptosis in diverse human cancer cell lines in which endogenous G0S2 was normally epigenetically silenced [38]. Although these results seem contradictory, it might also indicate that G0S2 has a dual role, on one side promoting cellular proliferation and on the other side inducing apoptosis under certain circumstances. Our results in ALK+ ALCL support a role  of G0S2 mainly in cell cycle progression with a minor role in apoptosis, potentially contributing to ALK-mediated oncogenesis.
Due to the emerging role of C/EBPb in the control of ribosomal RNA (rRNA) biogenesis during adipogenesis [42], it was remarkable to find that DDX21, previously known as RNA helicase II/Gua is a downstream target of C/EBPb. DDX21 (DEAD box polypeptide 21) belongs to the DExD/H box family of proteins that play important roles in basically all aspects of RNA metabolism [43]. Members of this family have been shown to act as RNA helicases with RNA unwinding and foldase activity, and are involved in the production of ribosomal RNA (rRNA). Interestingly, several DExD/H box proteins have been shown to be multifunctional and in addition to act as RNA helicases in processes such as pre-mRNA processing, translational initiation and ribosome assembly, play important roles in the regulation of gene transcription. DDX21 is a nucleolar protein with RNA helicase and folding activity involved in the regulation of rRNA biogenesis, it is critical for 18S and 28S rRNA biosynthesis and maintenance [27,28,44]. Loss of DDX21 results in blocking of rRNA biogenesis, slow metabolism and reduces cell proliferation [25,28]. In contrast high levels of DDX21 expression were shown to be associated with low survival and rapid relapse in breast cancer [45]. In previous studies, DDX21 was identified as a functional interacting partner for JUN in human cells [46], supporting JUN mediated target gene activation, whereas JUN regulated the rRNA binding and nucleolar localization of DDX21 [26]. The function of both proteins seems to be necessary to ensure cell growth. Accordingly, knockdown of DDX21 in SUDHL1 cells resulted in blocking of rRNA biogenesis, and a 65% reduction in cell proliferation compared to control cells. Annexin V/propidium iodide stainings after DDX21 knockdown indicated that DDX21 is also essential for survival of ALK+ ALCL cells. Production of rRNA is tightly coordinated with cell growth and proliferation, and reduced rRNA production after DDX21 silencing seems to result in decreased protein synthesis, slow metabolism, and inhibition of cell growth and proliferation.
In summary, using GEP analysis before and after C/EBPb silencing in ALK+ ALCL cells, we were able to identify several direct downstream targets of C/EBPb that might contribute to the ALK-mediated lymphoid transformation. Using the Ingenuity pathway tool, we constructed a model (Figure 7) to illustrate the important role of C/EBPb in different cellular pathways and its most important downstream targets. File S1 Table S1, S2, S3, S4. Table S1. Indicated primers and probes were combined using the Universal ProbeLibrary System for validation of gene expression regulation by C/EBPb of the 26 candidate genes by RT-qPCR. All primers were designed for intron-spanning multiplex assays with TBP. * intron-spanning not possible. Table S1: Indicated primers and probes were combined using the Universal ProbeLibrary System for validation of gene expression regulation by C/EBPb of the 26 candidate genes by RT-qPCR. All primers were designed for intron-spanning multiplex assays with TBP. * intron-spanning not possible. Table  S2: Primer sequences to amplify 49 promoter sequences of 11 genes with potential C/EBPb binding sites applying the QuantiTect Sybr Green PCR Kit. A BCL2A1 promoter sequence without potential C/EBPb binding site was amplified as negative control (BCL2A1_neg). Table S3: Shown are the 169 significant differentially expressed probe sets (FDR,10%; corresponding to 114 genes) regulated upon C/EBPb knockdown in the microarray analysis. Linear ratios are indicated (knockdown versus control). Table S4: Selected significantly (p,0.01) enriched GO terms and pathways associated with the set of 114 genes regulated upon C/ EBPb inhibition. (DOC)