Overexpression of a DEAD Box Protein (DDX1) in Neuroblastoma and Retinoblastoma Cell Lines*

The DEAD box gene, DDX1, is a putative RNA helicase that is co-amplified with MYCN in a subset of retinoblastoma (RB) and neuroblastoma (NB) tumors and cell lines. Although gene amplification usually involves hundreds to thousands of kilobase pairs of DNA, a number of studies suggest that co-amplified genes are only overexpressed if they provide a selective advantage to the cells in which they are amplified. Here, we further characterize DDX1 by identifying its putative transcription and translation initiation sites. We analyze DDX1 protein levels inMYCN/DDX1-amplified NB and RB cell lines using polyclonal antibodies specific to DDX1 and show that there is a good correlation with DDX1 gene copy number, DDX1 transcript levels, and DDX1 protein levels in all cell lines studied. DDX1 protein is found in both the nucleus and cytoplasm ofDDX1-amplified lines but is localized primarily to the nucleus of nonamplified cells. Our results indicate that DDX1 may be involved in either the formation or progression of a subset of NB and RB tumors and suggest that DDX1 normally plays a role in the metabolism of RNAs located in the nucleus of the cell.

The DEAD box gene, DDX1, is a putative RNA helicase that is co-amplified with MYCN in a subset of retinoblastoma (RB) and neuroblastoma (NB) tumors and cell lines. Although gene amplification usually involves hundreds to thousands of kilobase pairs of DNA, a number of studies suggest that co-amplified genes are only overexpressed if they provide a selective advantage to the cells in which they are amplified. Here, we further characterize DDX1 by identifying its putative transcription and translation initiation sites. We analyze DDX1 protein levels in MYCN/DDX1-amplified NB and RB cell lines using polyclonal antibodies specific to DDX1 and show that there is a good correlation with DDX1 gene copy number, DDX1 transcript levels, and DDX1 protein levels in all cell lines studied. DDX1 protein is found in both the nucleus and cytoplasm of DDX1-amplified lines but is localized primarily to the nucleus of nonamplified cells. Our results indicate that DDX1 may be involved in either the formation or progression of a subset of NB and RB tumors and suggest that DDX1 normally plays a role in the metabolism of RNAs located in the nucleus of the cell.
DEAD box proteins are a family of putative RNA helicases that are characterized by eight conserved amino acid motifs, one of which is the ATP hydrolysis motif containing the core amino acid sequence DEAD (Asp-Glu-Ala-Asp) (1)(2)(3). Over 40 members of the DEAD box family have been isolated from a variety of organisms including bacteria, yeast, insects, amphibians, mammals, and plants. The prototypic DEAD box protein is the translation initiation factor, eukaryotic initiation factor 4A, which, when combined with eukaryotic initiation factor 4B, unwinds double-stranded RNA (4). Other DEAD box proteins, such as p68, Vasa, and An3, can effectively and independently destabilize/unwind short RNA duplexes in vitro (5)(6)(7). Although some DEAD box proteins play general roles in cellular processes such as translation initiation (eukaryotic initiation factor 4A (4)), RNA splicing (PRP5, PRP28, and SPP81 in yeast (8 -10)), and ribosomal assembly (SrmB in Escherichia coli (11)), the function of most DEAD box proteins remains unknown. Many of the DEAD box proteins found in higher eukaryotes are tissue-or stage-specific. For example, PL10 mRNA is expressed only in the male germ line, and its product has been proposed to have a specific role in translational regulation during spermatogenesis (12). Vasa and ME31B are maternal proteins that may be involved in embryogenesis (13,14). p68, found in dividing cells (15), is believed to be required for the formation of nucleoli and may also have a function in the regulation of cell growth and division (16,17). Other DEAD box proteins are implicated in RNA degradation, mRNA stability, and RNA editing (18 -20).
The human DEAD box protein gene DDX1 1 was identified by differential screening of a cDNA library enriched in transcripts present in the two RB cell lines Y79 and RB522A (21). The longest DDX1 cDNA insert isolated from this library was 2.4 kb with an open reading frame from position 1 to 2201. All eight conserved motifs characteristic of DEAD box proteins are found in the predicted amino acid sequence of DDX1 as well as a region with homology to the heterogeneous nuclear ribonucleoprotein U, a protein believed to participate in the processing of heterogeneous nuclear RNA to mRNA (22,23). The region of homology to heterogeneous nuclear ribonucleoprotein U spans 128 amino acids and is located between the first two conserved DEAD box protein motifs, 1a and 1b.
The proto-oncogene MYCN encodes a member of the MYC family of transcription factors that bind to an E box element (CACGTG) when dimerized with the MAX protein (24,25). The MYCN gene is amplified and overexpressed in approximately one-third of all NB tumors (26,27). Amplification of MYCN is associated with rapid tumor progression and a poor clinical prognosis (26,27). MYCN overexpression is usually achieved by increasing gene copy number rather than by up-regulating basal expression of MYCN (27,28). Because gene amplification involves hundreds to thousands of kilobase pairs of contiguous DNA (29 -32), it is possible that co-amplification of a gene located in proximity to MYCN may contribute to the poor clinical prognosis of MYCN-amplified tumors. The DDX1 gene maps to the same chromosomal band as MYCN, 2p24, and is located ϳ400 kb telomeric to the MYCN gene (33)(34)(35)(36). All four MYCN-amplified RB tumor cell lines tested to date are amplified for DDX1 (21), 2 while approximately two-thirds of NB cell lines and 38 -68% of NB tumors are co-amplified for both genes (37)(38)(39). George et al. (39) found a significant decrease in the mean disease-free survival of patients with DDX1/MYCN-amplified NB tumors compared with MYCN-amplified tumors. Similarly, Squire et al. (38) observed a trend toward a worse clinical prognosis when both genes were amplified in the tumors of NB patients. To date, there have been no reports of a tumor amplified only for DDX1, and the role that this gene plays in cancer formation and progression is not known.
Because of the high rate of rearrangements in amplified DNA (31,40), it is unlikely that a gene located ϳ400 kb from the MYCN gene will be consistently amplified as an intact unit unless its product provides a growth advantage to the cell. Based on Southern blot analysis, the DDX1 gene extends over more than 30 kb, and there are no gross rearrangements of this gene in DDX1-amplified tumors (21,38). Furthermore, there is a good correlation between DDX1 transcript levels and gene copy number in the tumors analyzed to date. However, we need to show that DDX1 protein is overexpressed in DDX1-amplified tumors if we are to entertain the possibility that this protein plays a role in the tumorigenic process. Here, we isolate and characterize the 5Ј-end of DDX1 mRNA and extend the DDX1 cDNA sequence by ϳ300 nt. We identify the predicted initiation codon of DDX1 and generate antisera that specifically recognize DDX1 protein. We analyze levels of DDX1 protein in both DDX1-amplified and nonamplified RB and NB tumors and study the subcellular location of this protein in the cell.

MATERIALS AND METHODS
Library Screening-A human fetal brain cDNA library (Stratagene) was screened using a 320-bp DNA fragment from the 5Ј-end of the 2.4-kb DDX1 cDNA previously described (23). Phagemids containing positive inserts were excised from ZAP II following the supplier's directions. The ends of the cDNA inserts were sequenced using the dideoxynucleotide chain termination method with T7 DNA polymerase (Amersham Pharmacia Biotech).
A human placenta genomic library (CLONTECH) was screened with the 5Ј-end of DDX1 cDNA. Positive plaques were purified, and the genomic DNA was analyzed using restriction enzymes and Southern blotting. EcoRI-digested DNA fragments from these clones were subcloned into pBluescript and digested with exonuclease III and mung bean nuclease to obtain sequentially deleted clones. The exon/intron map of the 5Ј portion of the DDX1 gene was obtained by comparing the sequence of DDX1 cDNA with that of the genomic DNA.
Rapid Amplification of cDNA Ends (RACE)-We used the Ampli-FINDER RACE kit (CLONTECH) to extend the 5Ј-end of DDX1 cDNA. Briefly, two g of poly(A) ϩ RNA isolated from RB522A was reverse transcribed at 52°C using either primer P1 or P3 (Fig. 1A). The RNA template was hydrolyzed, and excess primer was removed. A singlestranded AmpliFINDER anchor containing an EcoRI site was ligated to the 3Ј-end of the cDNA using T4 RNA ligase. The cDNA was amplified using either primer P2 or P4 (Fig. 1A) and AmpliFINDER anchor primer. RACE products were cloned into pBluescript.
Primer Extension-Poly(A) ϩ RNAs were isolated from RB and NB cell lines as described previously (21,38). The 21-nt primers 5Ј-TTCGT-TCTGGGCACCATGTGT-3Ј (primer P4 in Fig. 1A) and 5Ј-TGGGAC-CTAGGGCTTCTGGAC-3Ј (primer P3 in Fig. 1A) were end-labeled with [␥-32 P]ATP (3000 Ci/mmol; Mandel Scientific) and T4 polynucleotide kinase. Each of the labeled primers was annealed to 2 g of poly(A) ϩ RNA at 45°C for 90 min, and the cDNA was extended at 42°C for 60 min using avian myeloblastosis virus reverse transcriptase (Promega). The primer extension products were heat-denatured and run on a 8% polyacrylamide gel containing 7 M urea in 1ϫ TBE buffer. A G ϩ A sequencing ladder served as the size standard. S1 Nuclease Protection Assay-The S1 nuclease protection assay to map the transcription initiation site of DDX1 was performed as described by Favaloro et al. (41). The DNA probe was prepared by digesting genomic DNA spanning the upstream region of DDX1 and exon 1 with AvaI, labeling the ends with [␥-32 P]ATP (3000 Ci/mmol) and polynucleotide kinase, and removing the label from one of the ends by digesting the DNA with SphI (Fig. 4). The RNA samples were resuspended in a hybridization mixture containing 80% formamide, 40 mM PIPES, 400 mM NaCl, 1 mM EDTA, and the heat-denatured SphI-AvaI probe labeled at the AvaI site. The samples were incubated at 45°C for 16 h and digested with 3000 units/ml S1 nuclease (Boehringer Mannheim) for 60 min at 37°C. The samples were precipitated with ethanol; resuspended in 80% formaldehyde, TBE buffer, 0.1% bromphenol blue, xylene cyanol; denatured at 90°C for 2 min; and electrophoresed in a 7 M urea, 8% polyacrylamide gel in TBE buffer.
Northern and Southern Blot Analysis-Poly(A) ϩ RNAs were isolated from RB and NB cell lines as described previously (21,38). Two g of poly(A) ϩ RNA/lane were electrophoresed in a 6% formaldehyde, 1.5% agarose gel in MOPS buffer (20 mM MOPS, 5 mM sodium acetate, 1 mM EDTA, pH 7.0) and transferred to nitrocellulose filter in 3 M sodium chloride, 0.3 M sodium citrate. The filters were hybridized to the following DNA probes, 32 P-labeled by nick translation: (i) a 1.6-kb EcoRI insert from DDX1 cDNA clone 1042 (21), (ii) a 260-bp cDNA fragment spanning the 3Ј-end of DDX1 exon 1 as well as exons 2 and 3, (iii) a 160-bp fragment derived from the 5Ј-end of DDX1 exon 1, and (iv) ␣-actin cDNA to control for lane to lane variation in RNA levels. Filters were hybridized and washed under high stringency. Southern blot analysis was as described previously (21).
Preparation of Anti-DDX1 Antiserum-To prepare antiserum to the C terminus of the DDX1 protein, we inserted a 1.8-kb EcoRI fragment from bp 848 to 2668 of DDX1 cDNA (Fig. 1B) into EcoRI-digested pMAL-c2 expression vector (New England Biolabs). DH5␣ cells transformed with this vector were grown to mid-log phase and induced with 0.1 mM isopropyl-1-thio-␤-D-thiogalactoside. The cells were harvested 3-4 h postinduction and lysed by sonication. Soluble maltose binding protein-DDX1 fusion protein was affinity-purified using amylose resin, and the maltose-binding protein was cleaved with factor Xa. The DDX1 protein was purified on a SDS-PAGE gel, electroeluted, and concentrated. Approximately 100 g of protein was injected into rabbits at 4 -6-week intervals. For the initial injection, the protein was dispersed in complete Freund's adjuvant (Sigma), while subsequent injections were prepared in Freund's incomplete adjuvant. Blood was collected from each rabbit 10 days after injection, and the specificity of the antiserum was tested using cell extracts from RB522A. To prepare antiserum to the N terminus of DDX1 protein, a DDX1 cDNA fragment from bp 268 to 851 ( Fig. 1B) was inserted into pGEX-4T2 (Amersham Pharmacia Biotech). The recombinant protein produced from this construct contains the first 186 amino acids of the predicted DDX1 sequence. Soluble glutathione S-transferase-DDX1 fusion protein was purified with glutathione-Sepharose 4B (Amersham Pharmacia Biotech). The glutathione S-transferase component of the fusion protein was cleaved with thrombin.
Subcellular Fractionations and Western Blot Analysis-We used two different procedures for subcellular fractionations. First, we isolated nuclear and S100 (soluble cytoplasmic) fractions from RB522A, IMR-32, Y79, RB(E)-2, HeLa, and HL60 using the procedure of Dignam (42). On average, we obtained 5-6 times more protein in the cytosolic fractions than in the nuclear fractions. Second, 10 8 RB522A cells were lysed and fractionated into S4 (soluble cytoplasmic components), P2 (heavy mitochondria, plasma membrane fragments), P3 (mitochondria, lysozymes, peroxisomes, and Golgi membranes), and P4 fractions (membrane vesicles from rough and smooth endoplasmic reticulum, Golgi, and plasma membrane) by differential centrifugation (43). We obtained 8 mg of protein in the S4 fraction, 1 mg in P2, 0.5 mg in P3, and 2 mg in P4 fraction. The procedures related to the immunoelectron microscopy have been previously described (44).
For Western blot analysis, proteins were electrophoresed in polyacrylamide-SDS gels and electroblotted onto nitrocellulose using the standard protocol for protein transfer described by Schleicher and Schuell. The filters were incubated with a 1:5000 dilution of DDX1 antiserum, a 1:200 dilution of anti-MYCN monoclonal antibody (Boehringer Mannheim), or a 1:200 dilution of anti-actin (Santa Cruz Biotechnology, Inc., Santa Cruz, CA). For the colorimetric analysis, antigen-antibody interactions were visualized using either alkaline phosphatase-linked goat anti-rabbit IgG (for DDX1) or goat anti-mouse IgG (for MYCN) at a 1:3000 dilution. For the ECL Western blotting analysis (Amersham Pharmacia Biotech), we used a 1:100,000 dilution of peroxidase-linked secondary anti-rabbit IgG antibody (for DDX1) or secondary anti-goat IgG antibody (Jackson ImmunoResearch Laboratories).

Identification of the 5Ј-End of the DDX1 Transcript-We
have previously reported the sequence of DDX1 cDNA isolated from an RB cDNA library (21,23). This 2.4-kb DDX1 cDNA contains an open reading frame spanning positions 1-2201 with a methionine encoded by the first three nucleotides (Fig.  1A). There is a polyadenylation signal and poly(A) tail in the 3Ј-untranslated region, indicating that the sequence is complete at the 3Ј-end. Manohar et al. (37) have also isolated DDX1 cDNA from the NB cell line LA-N-5. Their cDNA extended the 5Ј-end of our sequence by 42 bp and included an additional in frame methionine (double underlined in Fig. 1A). The possibil-ity of additional in frame methionines located further upstream could not be excluded, because there were no predicted stop codons in the upstream region of the cDNA.
Northern blot analysis indicated a DDX1 transcript size of ϳ2800 nt, suggesting that the DDX1 cDNAs isolated to date were lacking ϳ300 -350 bp of 5Ј sequence. We have used different approaches to identify the transcription start site of DDX1. First, we exhaustively screened a commercial fetal brain cDNA library with the 5Ј-end of DDX1 cDNA. Although numerous clones were analyzed, only one extended the sequence (by 35 bp) beyond that published by Manohar et al. (37) (Fig. 1A).
We next used the RACE procedure in an attempt to isolate additional 5Ј sequence. The nested primers used to amplify the 5Ј-end of the DDX1 transcript are labeled as primers P1 and P2 in Fig. 1A and are located downstream of the three in frame methionines (double underlined in Fig. 1A). Poly(A) ϩ RNA from RB522A was reverse transcribed at 52°C using primer P1, and the reverse transcribed cDNA was amplified using the nested primer P2 and the 5Ј-RACE primer. Using this approach, we generated a product that was 230 bp longer than any of the cDNAs obtained by screening libraries (Fig. 1A). Sequencing of this 230-bp cDNA revealed an in frame stop codon (boldface double underline in Fig. 1A) located 123 bp upstream of the predicted translation initiation site. We then prepared primers P3 and P4, located near the 5Ј-end of the RACE cDNA (Fig. 1A) and repeated the RACE procedure to see if additional 5Ј sequences could be obtained. The resulting RACE products did not extend the DDX1 cDNA sequence further.
The location of the DDX1 transcription initiation site was verified by primer extension. Poly(A) ϩ RNA was prepared from the following two cell lines: DDX1-amplified RB cell line RB522A and a nonamplified RB cell line RB(E)-2. RB522A has elevated levels of DDX1 mRNA, while RB(E)-2 has at least 20-fold lower levels of this transcript. Three products of 40, 43, and 46 nt (with a weak signal at 45 nt) were detected in RB522A using primer P4 (Figs. 1A and 2). The 40-nt product corresponded exactly with the 5Ј-end of the RACE-derived cDNA while the 43-and 46-nt products extended the predicted size of the DDX1 transcript by 3 and 6 nt, respectively. None of these products were observed in RB(E)-2. Bands of identical sizes to those obtained with RB522A mRNA were also observed in the DDX1-amplified NB cell line BE(2)-C but not in the DDX1-amplified NB cell line IMR-32 (data not shown). The same predicted DDX1 transcription initiation site was identified with primer P3 except that the bands were of weaker intensity (data not shown). We have designated the transcription start site identified by primer extension as ϩ1 (Fig. 1A).
The sequence of the 6 nt extending beyond the RACE cDNA was obtained by comparison of the cDNA sequence with that of DDX1 genomic DNA. Bacteriophages containing DDX1 genomic DNA were isolated by screening a human placenta library with 5Ј DDX1 cDNA. Eighteen kb of DNA were sequenced from two bacteriophages with overlapping DDX1 genomic DNA. Thirteen exons were identified within this 18-kb region (Fig. 3) corresponding to cDNA sequences from position 1 to 1249. The 310-bp exon 1 was by far the longest of the 13 exons sequenced, corresponding to the entire 5Ј-untranslated region of DDX1 as well as the first in frame methionine. The sequences transcribed from exons 1, 2, and 3 are indicated in Fig. 1A.
Knowledge of the genomic structure of DDX1 allowed us to use the S1 protection assay, a technique that is independent of reverse transcriptase, to further define the 5Ј-end of the  (Fig. 4). Specific bands were not detected in either DDX1-amplified Y79 and LA-N-5 or the nonamplified lines. Although the sizes of the S1 protected bands in RB522A, BE(2)-C, and LA-N-1 were 5 and 11 nt shorter than predicted based on RACE and primer extensions, respectively, there was general agreement with all three techniques regarding the location of the DDX1 transcription initiation site (Fig. 1A). The smaller S1 nuclease protected products could have arisen as the result of S1 digestion of the 5Ј-end of the RNA:DNA heteroduplex because of its relatively high rU:dA content (45).
Identification of the same transcription initiation site in three DDX1-amplified lines suggests that this represents the bona fide start site of DDX1 transcription. However, it was not clear why this start site was either very weak or not detected in three other amplified lines. To determine whether the 5Ј-end of exon 1 is transcribed in all DDX1-amplified lines, we carried out a direct analysis of the 5Ј-end of the DDX1 transcript by Northern blotting. Two probes were used for this analysis: the 5Ј probe contained a 160-bp fragment from bp 1 to 160 (5Ј-half of exon 1), and the 3Ј probe contained a 260-bp fragment from bp 160 to 420 (3Ј-half of exon 1 as well exons 2 and 3) (Fig. 1A). With the 3Ј probe, we obtained bands of similar size and intensity in four DDX1-amplified lines (RB522A, BE(2)-C, IMR-32, and LA-N-5). Band intensity was somewhat weaker in Y79 and stronger in LA-N-1 in comparison with the other lines (Fig. 5). No signal was detected in the non-DDX1-amplified line RB412. With the 5Ј probe, a relatively strong signal was observed in RB522A, BE(2)-C, and LA-N-1, while a considerably weaker but readily apparent signal was detected in Y79, IMR-32, and LA-N-5. The signal obtained with actin indicates that, with the exception of LA-N-1, similar amounts of RNA were loaded in each lane and that the RNA was not degraded. These results indicate that at least a portion of the 160-bp 5Ј-end of exon 1 is transcribed in all DDX1-amplified lines.
Based on primer extension, S1 nuclease protection assay, Northern blot analysis and the sequencing of the RACE products, we conclude that the DDX1 transcript is 2.7 kb with an open reading frame spanning nucleotides 295-2515 encoding a predicted protein of 740 amino acids with an estimated molecular weight of 82.4 (Fig. 1B). An in frame stop codon is located 123 nt upstream of the predicted translation initiation site, at positions 172-174. The first in frame methionine following the stop codon is in agreement with the Kozak consensus sequence (46). Furthermore, the predicted start methionine codon for human DDX1 corresponds perfectly with that of Drosophila DDX1 (47). A stop codon is located 15 nt upstream of the initiation codon in Drosophila DDX1.
Analysis of DDX1 Protein Levels in Neuroblastoma and Retinoblastoma-We and others have previously shown that there is a good correlation between gene copy number and RNA levels in DDX1-amplified RB and NB cell lines (37,38). To determine whether the correlation extends to DDX1 protein levels, we prepared antiserum to two nonoverlapping recombinant DDX1 proteins. First, we prepared a C terminus recombinant protein construct by inserting a 1.8-kb EcoRI fragment from bp 848 to 2668 (amino acids 185-740) (Fig. 1B) into the pMAL-c2 expression vector. Recombinant protein expression was induced with isopropyl-1-thio-␤-D-thiogalactoside, and the 110-kDa maltosebinding protein-DDX1 fusion product was purified by affinity chromatography using amylose resin, followed by electrophoresis on a SDS-PAGE gel after cleaving the maltose-binding protein fusion partner with factor Xa. Second, we prepared an N terminus construct by ligating a DNA fragment from bp 268 to 851 (amino acids 1-186) into pGEX-4T2. The 50-kDa glutathione S-transferase-DDX1 fusion protein was purified by affinity chromatography on a glutathione column. This N terminus fusion protein contains only the first of the eight conserved motifs found in all DEAD box proteins, while the C terminus fusion protein includes the remaining seven motifs.
We measured DDX1 protein levels in total cell extracts of three RB and 10 NB cell lines. Using antiserum to the N terminus fusion protein, we observed a strong signal in all  (Fig. 6). Two bands were observed in the majority of extracts. Of the amplified lines, Y79 produced the weakest signal, with the most intense signal observed in LA-N-1. There was an excellent correlation with DDX1 protein and mRNA levels in these cell lines, with lower levels of DDX1 mRNA observed in Y79 and higher levels in LA-N-1 (Fig. 7A). As shown in Fig. 7B, this correlation extended to DDX1 gene copy number. No gross DNA rearrangements were seen in the DDX1-amplified lines; however, three small bands of altered size were observed in the RB412 lane. Although the nature of the DNA alteration is not known, it is noteworthy that DDX1 transcript levels in RB412 are extremely low (Fig. 7A) and that the top DDX1 protein band in RB412 cell extracts is smaller in size than the top band from the other cell extracts (Fig. 6).
Two DDX1 protein bands were present in most of the lanes in Fig. 6. The same two bands were detected with antiserum to the C terminus of the DDX1 protein, as well as a third band at ϳ60 kDa (data not shown). There was no variation in the intensity of the 60-kDa band in DDX1-amplified and nonamplified cell extracts. The 60-kDa band probably represents another member of the DEAD box protein family, because the C terminus DDX1 protein used to prepare this antiserum contained seven of the eight conserved motifs found in all DEAD box proteins. To obtain an estimate of the size of the two DDX1 bands, we ran cellular extracts from RB522A on a 7% SDS-PAGE gel with the BenchMark protein ladder (Life Technologies, Inc.). The size of the DDX1 protein was determined using the Alpha Imager 2000 documentation and analysis system for molecular weight calculation. Based on this analysis, the estimated molecular mass of the top band is 89.5 kDa, while that of the bottom band is 83.5 kDa. The 84-kDa band may represent the unmodified product encoded by the DDX1 transcript (capable of encoding a protein with a predicted molecular mass of 82.4 kDa), while the top band may represent post-translational modification of DDX1 protein (e.g. phosphorylation). Another possibility is that the top band represents intact DDX1 and the lower band is a specific truncated or degradation product of DDX1. Yet a third possibility is that the two bands represent the products of differentially spliced transcripts or different translation initiation sites. However, the lack of any obvious differences in DDX1 transcript sizes in the three RB and 10 NB lines analyzed in Fig. 7A does not support the latter possibility (e.g. compare the DDX1 transcript size in NUB-7 (which produces the lower DDX1 protein band) and in NB-1 (which produces the higher DDX1 protein band)).
Subcellular Localization of DDX1 Protein-DEAD box proteins have been implicated in a variety of cellular functions including RNA splicing in the nucleus, translation initiation in the cytoplasm, and ribosome assembly in the nucleolus. To obtain an indication of the possible role of DDX1, we studied its subcellular location. Nuclear and cytosolic extracts were prepared from DDX1-amplified RB522A and run on a 7% SDS-PAGE gel. Although there was more DDX1 protein in the cytosol than in the nucleus on a per cell basis, the proportion of DDX1 protein relative to total protein was similar in both cellular compartments (Fig. 8A). Both the 90-and 84-kDa bands were present in cytosol and nuclear extracts, although the bottom band was more readily apparent in the cytosol. By running the gel for an extended period of time (twice as long as usual), we were able to detect an additional weak band at ϳ88 kDa in both nuclear and cytosolic extracts.
To determine whether DDX1 consistently localizes to both the cytoplasm and nucleus, we prepared cytosol and nuclear extracts from two additional DDX1-amplified lines, Y79 and IMR-32, as well as from nonamplified RB(E)-2, HL60, and HeLa. DDX1 protein was found in both the nucleus and cytoplasm of IMR-32, primarily in the cytoplasm of Y79, and mainly in the nucleus of the three nonamplified lines (Fig. 8B). In addition, DDX1 was almost exclusively found in nuclear  were loaded in each lane, electrophoresed in a 1.5% agarose-formaldehyde gel, and transferred to a nitrocellulose filter. The filter was first hybridized to a 32 P-labeled 1.6-kb DDX1 cDNA (clone 1042) (21), stripped, and rehybridized to actin DNA. B, 10 g of genomic DNA from each of the indicated cell lines were digested with EcoRI, electrophoresed in a 1% agarose gel, and transferred to a nitrocellulose filter. The filter was hybridized to 32 P-labeled clone 1042 DDX1 cDNA, stripped, and reprobed with labeled ␣-fetoprotein cDNA. Markers (in kilobase pairs) are indicated on the side.

FIG. 8. Distribution of DDX1 in the nucleus and cytoplasm.
A, cytosolic and nuclear extracts were prepared from RB522A and electrophoresed in a 7% SDS-PAGE gel. Cytosolic extracts were loaded in lanes 1 (20 g of protein) and 2 (10 g), while nuclear extracts were loaded in lanes 3 (10 g) and 4 (20 g). DDX1 was visualized using a 1:5000 dilution of the antiserum to the N terminus. The BenchMark protein ladder size markers (kilodaltons) are indicated on the left. B, cytosolic and nuclear extracts were prepared from HL60, Y79, IMR-32, HeLa, RB522A, and RB(E)-2 and electrophoresed in an 8% SDS-PAGE gel. Twenty g of proteins were loaded in each lane marked C (cytosolic) and N (nuclear). DDX1 was visualized using a 1:5000 dilution of the antiserum to the N terminus. Actin levels were analyzed using a 1:200 dilution of anti-actin antibody (Santa Cruz Biotechnology). extracts prepared from normal GM38 fibroblasts (data not shown). We used anti-actin antibody to ensure that our nuclear and cytosolic extracts were not cross-contaminated (Fig. 8B).
We next carried out a more detailed analysis of DDX1 subcellular location using two different approaches: (i) fractionation of cellular components into nuclei; S100 or S4 cytosol (containing soluble cytoplasmic components, including 40 S ribosomes); P2 (heavy mitochondria, plasma membrane fragments plus material trapped by these membranes); P3 (mitochondria, lysosomes, peroxisomes, Golgi membranes, some rough endoplasmic reticulum); and P4 (microsomes from smooth and rough endoplasmic reticulum, Golgi and plasma membranes) (43); and (ii) immunogold electron microscopy. The DDX1-amplified RB522A cell line was used for both experiments. The fractionation procedures indicate that DDX1 is mainly in the nucleus and in the cytosol (S4 and S100 fractions) of RB522A cells (Fig. 9A). As a control, we used anti-human MYCN antibody to determine the location of MYCN (also amplified in RB522A) in our subcellular fractions. As shown in Fig. 9B, MYCN was primarily found in the nucleus, as one would expect of a transcription factor.
For the electron microscopy analysis, antiserum to the N terminus of DDX1 was coupled to protein A gold particles, and the distribution of DDX1 was examined in RB522A cells fixed in paraformaldehyde and glutaraldehyde. DDX1 was present in both the cytoplasm and nucleus (data not shown). There was no association with either cell organelles or with nuclear or plasma membranes. DISCUSSION There are presently few clues as to the function of DDX1 in normal and cancer cells. Our earlier data indicate that DDX1 mRNA is present at higher levels in fetal tissues of neural origin (retina and brain) compared with other fetal tissues (21).
There may therefore be a requirement for elevated levels of this putative RNA helicase for the efficient production or processing of neural specific transcripts. A role in cancer formation or progression is an intriguing possibility, because overexpression of an RNA unwinding protein could affect the secondary structure of RNAs in such a way as to alter the expression of specific proteins in tumor cells. DDX1 is co-amplified with MYCN in a subset of RB and NB cell lines and tumors (37)(38)(39). MYCN amplification is common in stage IV NB tumors and is a well documented indicator of poor prognosis. A general trend toward a poorer clinical prognosis is observed when both the MYCN and DDX1 genes are amplified compared with when only MYCN is amplified (38,39), suggesting a possible role for DDX1 in NB tumor formation or progression.
It is generally accepted that co-amplified genes are not overexpressed unless they provide a selective growth advantage to the cell (48,49). For example, although ERBA is closely linked to ERBB2 in breast cancer and both genes are commonly amplified in these tumors, ERBA is not overexpressed (48). Similarly, three genes mapping to 12q13-14 (CDK4, SAS, and MDM2) are overexpressed in a high percentage of malignant gliomas showing amplification of this chromosomal region, while other genes mapping to this region (GADD153, GLI, and A2MR) are rarely overexpressed in gene-amplified malignant gliomas (50,51). The first three genes are probably the main targets of the amplification process, while the latter three genes are probably incidentally included in the amplicons. The data shown here indicate that DDX1 is overexpressed at both the protein and RNA levels in DDX1-amplified RB and NB cell lines and that there is a strong correlation between DDX1 gene copy number, DDX1 RNA levels, and DDX1 protein levels in these lines. Our results are therefore consistent with DDX1 overexpression playing a positive role in some aspect of NB and RB tumor formation or progression. Recently, Weiss et al. (52) have shown that transgenic mice that overexpress MYCN develop NB tumors several months after birth. They conclude that MYCN overexpression can contribute to the initiation of tumorigenesis but that additional events are required for tumor formation. Amplification of DDX1 may represent one of many alternative pathways by which a normal precursor "neuroblast" or "retinoblast" cell gains malignant properties.
The function of the majority of tissue-specific or developmentally regulated DEAD box genes remains unknown. However, some members of this protein family have been either directly or indirectly implicated in tumorigenesis. For example, the p68 gene has been found to be mutated in the ultraviolet lightinduced murine tumor 8101 (53), while DDX6 (also known as RCK or p54) is encoded by a gene located at the breakpoint of the translocation involving chromosomes 11 and 14 in a cell line derived from a B-cell lymphoma (54,55). Similarly, the production of a chimeric protein between DDX10 and the nucleoporin gene NUP98 has been proposed to be involved in the pathogenesis of a subset of myeloid malignancies with inv(11) (p15q22) (56). Interestingly, Grandori et al. (57) have shown that MYCC interacts with a DEAD box gene called MrDb, suggesting that the transcription of some DEAD box genes could be regulated through interaction with members of the MYC family. Future work will involve determining whether DDX1 represents another member of the DEAD box family with a role in the tumorigenic process.
DEAD box proteins have been implicated in translation initiation, RNA splicing, RNA degradation, and RNA stability (3,18,19). We carried out subcellular localization studies in an attempt to obtain a general indication of the function of DDX1. We found DDX1 protein in both the cytoplasm and nucleus of DDX1-amplified NB and RB lines. In contrast, DDX1 was mainly located in the nucleus of nonamplified cell lines and normal fibroblast cultures. DDX1 was not associated with cellular organelles or with membranes based on immunoelectron microscopy. We therefore propose that the primary role of DDX1 is in the nucleus. The presence of DDX1 in the cytoplasm of DDX1-amplified cells may indicate that the amount of DDX1 protein that is allowed in the nucleus is tightly regulated. Alternatively, DDX1 may play a dual role in the nucleus and cytoplasm of DDX1-amplified cells.
An important component of our analysis was to identify the translation and transcription initiation sites of DDX1. We used a combination of techniques to identify the transcription start site: screening of RB and fetal brain libraries, RACE, primer extension, genomic DNA sequencing, S1 nuclease mapping, and Northern blot analysis using probes to the predicted 5Ј-end of the transcript. The transcription start site identified using these techniques is located ϳ300 nt upstream of the predicted translation initiation codon and was readily detected in three DDX1-amplified lines and barely detectable in a fourth amplified line. The 5Ј-untranslated region as well as the first in frame methionine are encoded within the first exon of DDX1. An in frame stop codon is located 123 nt upstream of the predicted initiation codon. We were unable to identify the transcription initiation site of DDX1 in two of the six amplified lines tested as well as in nonamplified lines. Although it remains possible that there are different transcription start sites in different cell lines, detection of lower levels (rather than the absence) of the 5Ј-most 160 nt of the DDX1 transcript in IMR-32, Y79, and LA-N-5 compared with RB522A, BE(2)-C, and LA-N-1 supports a quantitative rather than a qualitative difference in the 5Ј-end of this transcript in these cells. Our results suggest that the 5Ј-end of DDX1 mRNA is rarely intact, even in mRNA preparations that otherwise appear to be of high quality based on analysis of control transcripts. The 5Ј-end of DDX1 mRNA may therefore be especially susceptible to degradation, perhaps because of its sequence and/or secondary structure.
In conclusion, we have mapped the 5Ј-end of the 2.7-kb DDX1 transcript and have identified the predicted translation initiation site of DDX1 protein. We have found that DDX1-amplified RB and NB tumor lines overexpress DDX1 protein and that there is a good correlation between gene copy number and both transcript and protein levels in these cells. We have shown that DDX1 protein is primarily located in the nucleus of cells that are not DDX1-amplified. In contrast, DDX1 is present in both the nucleus and cytoplasm of DDX1-amplified NB and RB lines. A cytoplasmic location in DDX1-amplified lines may indicate that the amount of nuclear DDX1 is tightly regulated or that DDX1 plays a dual role in the cytoplasm and nucleus of these cells.