A Constitutional Translocation t(1;17)(p36.2;q11.2) in a Neuroblastoma Patient Disrupts the Human NBPF1 and ACCN1 Genes

The human 1p36 region is deleted in many different types of tumors, and so it probably harbors one or more tumor suppressor genes. In a Belgian neuroblastoma patient, a constitutional balanced translocation t(1;17)(p36.2;q11.2) may have led to the development of the tumor by disrupting or activating a gene. Here, we report the cloning of both translocation breakpoints and the identification of a novel gene that is disrupted by this translocation. This gene, named NBPF1 for Neuroblastoma BreakPoint Family member 1, belongs to a recently described gene family encoding highly similar proteins, the functions of which are unknown. The translocation truncates NBPF1 and gives rise to two chimeric transcripts of NBPF1 sequences fused to sequences derived from chromosome 17. On chromosome 17, the translocation disrupts one of the isoforms of ACCN1, a potential glioma tumor suppressor gene. Expression of the NBPF family in neuroblastoma cell lines is highly variable, but it is decreased in cell lines that have a deletion of chromosome 1p. More importantly, expression profiling of the NBPF1 gene showed that its expression is significantly lower in cell lines with heterozygous NBPF1 loss than in cell lines with a normal 1p chromosome. Meta-analysis of the expression of NBPF and ACCN1 in neuroblastoma tumors indicates a role for the NBPF genes and for ACCN1 in tumor aggressiveness. Additionally, DLD1 cells with inducible NBPF1 expression showed a marked decrease of clonal growth in a soft agar assay. The disruption of both NBPF1 and ACCN1 genes in this neuroblastoma patient indicates that these genes might suppress development of neuroblastoma and possibly other tumor types.


Introduction
Neuroblastoma, a tumor derived from pluripotent neuroblasts, is the most frequent extracranial solid malignancy of childhood. It is remarkably heterogeneous both clinically and biologically, as there is a large variation in tumor progression depending on the age of the patient at the time of diagnosis and the tumor stage [1].
Different genetic abnormalities are frequently observed in primary neuroblastoma tumors and their derivative cell lines. The most frequent aberration is the unbalanced gain of 17q, often due to translocation of the 17q segment to other chromosomes, combined with retention of two copies of the normal chromosome 17. One of the most common translocation partners for the 17q fragment is chromosome 1p, indicating that this translocation confers a selective advantage on cells by disrupting or activating certain genes [2,3]. As the translocation breakpoints are scattered on chromosomes 1 and 17, there is little evidence for a genespecific inactivation [4].
Amplification of the proto-oncogene MYCN, which occurs in about one quarter of all neuroblastomas [1], is correlated with poor prognosis [5]. Several studies have shown a strong correlation between MYCN amplification and deletion of the short arm of chromosome 1 [6], another frequent abnormality in neuroblastomas. Other deletions frequently affect chromosomes 3p and 11q [1] and hint at the presence of tumor suppressor genes (TSGs) in these chromosomal regions.
Previously, we described a de novo, apparently balanced, constitutional translocation t(1;17)(p36.2;q11.2) in a neuroblastoma patient [7]. According to Knudson's two-hit hypothesis, constitutional deletions and translocations in patients with specific tumors are often the first indicators of the presence of TSGs. Several lines of evidence lead to the hypothesis that the 1p36 locus harbors one or more TSGs for neuroblastoma. Introduction of the short arm of chromosome 1 in the neuroblastoma cell line NGP results in a differentiated phenotype [8], providing experimental evidence for this hypothesis. Furthermore, constitutional deletions or translocations affecting the 1p36 region have been described in other neuroblastoma patients [9,10]. Several other tumor types, such as breast cancer, malignant melanoma and colorectal cancer, are also characterized by frequent deletions or translocations of 1p36 [11]. In several types of tumors, including neuroblastoma, the shortest regions of overlap for 1p deletions partly coincide, indicating that the same TSGs may be implicated in different types of tumors [12]. A number of strong candidate TSGs were recently reported for 1p, including the CHD5 gene [13] and the microRNA-34a [14]. However, deletions in chromosome 1p36 are often very large in neuroblastoma [15], and it is likely that defects in more than one gene may contribute to the malignant phenotype. Expression profiling showed that expression levels of some genes located in the 1p35-36 region were decreased in neuroblastoma tumors and cell lines with 1p deletion as compared to 1p-normal samples [16,17]. Consequently, it was proposed that the underlying reason for the frequent deletion of this region in neuroblastoma is the decreased expression of several TSGs located in the 1p36 region and not the inactivation of one single classical TSG.

Probe generation
We previously described a cosmid contig spanning the chromosome 17 breakpoint [18]. Nine regions lacking repetitive elements were chosen as probes for Southern analysis and amplified by PCR using PAC RPCI-1 880L8 as template. Probe 6, which is localized on the cosmid contig (GenBank Acc Nu AF148647) from position 36,914 to 37,391, was amplified with primers 59-TCCCAAAAGGCCAGTTTCACAC-39 and 59-TCTGCAGGCGTCTCATCTCAAC-39. Probe 9 is located from position 38,018 to 38,584, and was amplified with primers 59-GTTGTACCCCCTTGACTTCA-39 and 59-GTGCCCAG-CAGGAGATTCAAT-39.

Southern blotting
Genomic DNA from normal human placenta, the Chinese hamster ovary cell line A/3, and somatic cell hybrids 32-7A and 32-2F53VIII [7] was extracted using standard procedures. DNA was digested with BglII, HindIII, SacI, XhoI, XbaI, McoI, PvuII, KpnI or HincII. Blotting and hybridization were done according to standard procedures.

GenomeWalker Libraries and PCR amplification
GenomeWalker libraries were generated according to the manufacturer's instructions (BD Biosciences, Palo Alto, CA, USA). For the 32-2F53VIII cell line, 2.5 mg of genomic DNA was digested with PvuII, DraI, StuI or EcoRV. For the 32-7A cell line, 2 mg of genomic DNA was digested with DraI, EcoRV, PvuII, ScaI or SspI. Subsequently, the appropriate GenomeWalker adaptors were ligated to the DNA.

RACE analysis
39RACE was performed with a commercial kit according to the manufacturer's instructions (Invitrogen, Merelbeke, Belgium).
To search for chimeric ACCN1 transcripts, we performed 59 and 39RACE analysis on cDNA from the 32-2F53VIII and 32-7A cell lines, respectively, using the Generacer kit according to the manufacturer's instructions (Invitrogen). Briefly, 2 mg of total RNA was reverse transcribed using the supplied oligo-dT primer. In the 39RACE experiment, we performed a PCR with a GSP (59-AGCACTACAAACCCAAGCAGTTC-39) in combination with the supplied 39 adaptor primer. Nested PCR was done using a GSP (59-GTGGGCCATGACCTGAAGGATAT-39) in combination with the supplied 39 nested primer. In the 59RACE experiment, we used a GSP (59-CTTGGATGAAAGGTGGCT-CAGACT-39) together with the supplied 59 adaptor primer. Nested PCR was performed using a GSP (59-TGATCTC-CAGCCCGTTGCCTGTC-39) and the supplied 59 nested primer. Products were cloned in pGEM-T Easy (Promega, Madison, WI, USA).

mRNA expression profiling by real-time quantitative PCR
For the analysis of the global NBPF expression pattern, we determined relative gene expression levels using an optimized two-step SYBR Green I RT-PCR assay employing several reference genes [19]. After establishment of amplification efficiencies of .95% for all primer pairs, the delta-Ct method was used for quantification. SYBR Green I core PCR reagents were obtained from Eurogentec (Seraing, Belgium) and used according to the manufacturer's instructions. Reactions were run on an iCycler (Bio-Rad, California, USA). Gene expression levels were normalized using the geometric mean of four reference genes [19]. Relative expression levels for amplicons in the NBPF transcripts were measured in a panel of 31 neuroblastoma cell lines.
For analysis of the NBPF1 expression pattern we determined the relative gene expression levels by an optimized two-step SYBR Green I RT-PCR assay using the two most stable housekeeping genes in neuroblastoma cell lines [19]. One microgram of total RNA was treated with the RQ1 RNase-free DNase according to the manufacturer's instructions (Promega). Before synthesizing cDNA, treated RNA samples were desalted on Microcon-100 spin columns (Milipore, Bedford, USA). cDNA was synthesized at 54uC using gene-specific primers (NBPF1: 59-CTGGGTGGACTTCGCGTAACTT-39, GAPD: 59-GGCATG-GACTGTGGTCATGAG-39 and HPRT1: 59-GGTCCTTTTC-ACCAGCAAGCT-39) and Superscript III reverse transcriptase according to the manufacturer's instructions (Invitrogen). RT-PCR mixes contained 20 ng template cDNA, LightCycler 480 SYBR Green I Mastermix (Roche Diagnostics GmbH, Mannheim, Germany) and 900 nM forward and reverse primers. Reactions were performed on a LightCycler 480 (Roche Diagnostics) using the following protocol: incubation at 95uC for 5 min, then 50 cycles at 95uC for 10 sec, 66uC for 30 sec, and 72uC for 1 sec.
Primer sequences for the reference genes are deposited in RTPrimerDB, a public database for real-time PCR primers (http://medgen.ugent.be/rtprimerdb/; [20]): UBC, SDHA, HPRT1, and GAPD. The primer sequences for the NBPF All measurements were performed in triplicates and the mean C p -value was used in the calculations. Statistical analysis of the groups with NBPF1 loss and those without loss was performed with the Wilcoxon rank sum test.

Cell culture and transfection
The colon cancer cell line DLD1Tr21 Tet-On was obtained from Van de Wetering et al. [21]. This Tet-On system activates transcription of the gene of interest in the presence of doxycycline (Dox). The cells were cultured in RPMI with 10% fetal calf serum, 100 U/ml penicillin and 100 mg/ml streptomycin. The cDNA for NBPF1-IRES-GFP, fused to an aminoterminal flag tag, was cloned in the pcDNA4/TO vector (Invitrogen). This construct was stably transfected in DLD1Tr21 cells using Lipofectamine reagent (Invitrogen) according to the manufacturer's instructions. Fluorescence activated cell sorting allowed rapid isolation of NBPF1-positive cells, and further subcloning of this population yielded a clone that was more than 90% positive for NBPF1 expression upon induction with Dox (2 mg/ml, Sigma, Steinheim, Germany) for 48 h. As the transfected vector encodes a flag-tagged NBPF1 fusion protein, single colonies were tested for induction of NBPF1 expression by immunofluorescence staining with the anti-flag antibody (Sigma).

Colony formation in soft agar
Soft agar assays were performed with 6,500 DLDTr21 cells, either stably transfected with a GFP-expressing vector or with the NBPF1-IRES-GFP-expressing vector, both in the presence and absence of Dox. The cells were cultured in six-well dishes in a two-layer soft agar system, with 0.7% (w/v) Bacto agar (Difco, Detroit, MI, USA) on the bottom and 0.35% (w/v) Bacto agar on the top. Liquid medium was placed on top of the agar. During incubation, this medium was refreshed every 2 days to provide a continuous source of Dox. An additional top-agar overlay was given after one week. Experiments were performed in duplicates and the mean value was used for further analysis. Plates were incubated at 37uC with a humidified atmosphere of 5% CO 2 for 2 weeks and colonies were counted using an inverted microscope. Pictures are taken with a Leica microscope (0.86magnification) or with an Olympus microscope (46magnification).

Ethics
The reported procedures using human samples were fully approved by the Ethics Committee at Ghent University and written informed consent was obtained from each patient since 1983. For patients whose DNA was obtained before 1983, only verbal consent was obtained. Indeed, the Ethics Committee advised that oral consent was sufficient for those patients, and that written consent should not be sought because most of these patients had died, and it was considered more ethical not to raise the matter with the family members again.

Identification of rearranged genomic fragments encompassing the constitutional t(1;17) translocation breakpoint
Karyotypic analysis of the lymphocytes of a neuroblastoma patient revealed a de novo, constitutional, apparently balanced translocation between chromosomes 1p36.2 and 17q11.2 [7]. Karyotyping of the primary tumor was unsuccessful, mostly due to necrosis. Due to the highly repetitive structure of the chromosome 1p36 region, breakpoint cloning proved to be extremely difficult. For this reason, we constructed a PAC/BAC contig of 1.5 Mbp covering the 17q breakpoint, and by screening a chromosome-17 specific cosmid library we identified three cosmids overlapping the breakpoint [18].
To facilitate the identification and cloning of the t(1;17) translocation breakpoints, genomic DNA from the somatic cell hybrids 32-7A and 32-2F53VIII, the Chinese hamster ovary cell line A/3, and normal human placenta were used for Southern blot analysis. The somatic cell hybrids 32-7A and 32-2F53VIII contain, respectively, the der(1) and the der (17) chromosomes of this neuroblastoma patient [7]. Based on the sequence of the cosmid contig spanning the chromosome 17 translocation breakpoint (GenBank Acc Nu AF148647), nine DNA probes free of repetitive sequences were selected ( Figure 1A). These probes were used for Southern blot analysis to search for abnormal, patient-specific hybridization bands. Probes 6 and 9 revealed additional bands in the DNA of cell lines 32-2F53VIII and 32-7A, respectively, but not in any of the other DNAs analyzed, indicating that these fragments span the chromosome 17 translocation breakpoint ( Figure 1B). Both hybrid cell lines also contain a copy of the normal chromosome 17, resulting in the appearance of a 'normal' band in these cell lines.
Cloning from 32-2F53VIII. Primer pair GSP1 and AP1 was used in a first round of amplification. Specific products of 1,457 bp and 3,575 bp were amplified from, respectively, the 32-2F53VIII/StuI and 32-2F53VIII/PvuII GenomeWalker libraries ( Figure 1C). With the GSP2 and AP1 primer pair, products of 1,029 bp and 3,459 bp were amplified from, respectively, the 32-2F53VIII/DraI and 32-2F53VIII/EcoRV GenomeWalker libraries. All four 32-2F53VIII-specific products were cloned, and complete sequencing showed overlapping novel sequences derived from chromosome 1 in all clones. In total, we sequenced 3,743 bp flanking the breakpoint region on the der(17) chromosome, of which 3,134 nucleotides were derived from chromosome 1 (GenBank Acc Nu AF379607; partly shown in Figure 2A).
Primer pair GSP4 and AP2 amplified products of 1,268 bp and 4,355 bp from, respectively, the 32-7A/SspI and 32-7A/PvuII GenomeWalker libraries. All four 32-7A-specific products were cloned, and complete sequencing showed overlapping novel sequences derived from chromosome 1 in all clones. In total, we sequenced 4,512 nucleotides, of which 3,845 bp were derived from chromosome 1 (GenBank Acc Nu AF379606; partly shown in Figure 2B). Previous FISH experiments indicated that the constitutional translocation was balanced [7]. To verify this at the molecular level, we compared the chromosome 17 sequences obtained by GenomeWalker PCR to the sequence of the cosmid contig spanning the der(17) translocation breakpoint. We found that the translocation was indeed balanced on this level ( Figure 2C). However, comparison of the chromosome 1 sequences to the human genome reference sequence (NCBI Build 36.1) revealed greater complexity.
In both derivative chromosomes, a few nucleotides of unknown origin had been inserted between the chromosome 1 and chromosome 17 sequences (Figures 1 and 2), a feature frequently observed in reciprocal constitutional translocations [22]. Remarkably, both the chromosome 1 translocation breakpoint cloned from the der(1) chromosome and the chromosome 17 breakpoint were located in LINE repeats ( Figure 3). The repeat on chromosome 1 had been classified as type L1-PA4 [23], and the LINE repeat on chromosome 17 as type L2. However, when we compared these two LINE repeats, no significant sequence homology was found. On the other hand, the chromosome 1 breakpoint cloned from the der(17) chromosome was located in a repeat-free region.
We previously reported that ACCN1 is the closest gene to the 17q breakpoint [18], but after completion of the human genome sequence it became clear that the translocation disrupts one of the isoforms of ACCN1 (Figure 3). This gene encodes an amiloridesensitive cation channel [24] with two transmembrane domains [25]. It was shown that the ACCN1 gene is expressed as two isoforms differing in their 59UTR and N-terminal domain (Figure 3) [26]. All exons encoding isoform 1 (GenBank Acc Nu NM_183377.1) are located proximal to the translocation breakpoint, whereas the first exon of isoform 2 (GenBank Acc Nu NM_001094.4) is located distal to the breakpoint. The remaining exons of isoform 2 are shared with isoform 1 and are proximal to the breakpoint. The translocation thus disrupts isoform 2 of the ACCN1 gene without disturbing isoform 1 (Figure 3). Both isoforms are also found in mouse and rat, strengthening their functional relevance. On the translocated 1p-fragment of the der (17) chromosome, we identified three exons encoding the carboxyterminal domain of the recently described NBPF1 protein. These are shown as exons of types 11, 12 and 14 in figures 1C and 3 (see [27] for numbering of NBPF1 exon types). NBPF1 is a member of a novel gene family with an intricate genomic structure [27][28][29]. Analysis of the der(1) sequences proximal to the breakpoint revealed the presence of two exons (shown as exons of types 6 and 7 in Figure 1D and Figure 3) that encode an internal domain of the NBPF1 protein. From this we conclude that the two breakpoints are located within distinct introns of the NBPF1 gene.
Although the translocation appeared cytogenetically balanced, molecular cloning of the breakpoint sequences revealed a deletion of part of the NBPF1 gene. Recent reports describing structural variation in the human genome have shown recurrent structural abnormalities in the NBPF1 locus in normal individuals [30,31]; we also observed these abnormalities in our analysis of NBPF1 cDNA clones [27]. Due to this diversity, it is not possible to precisely determine the length of the deleted genomic fragment on chromosome 1 in this patient. However, compared to the human genome reference sequence (NCBI Build 36.1), a fragment comprising at least four exons of the NBPF1 gene is missing (5,215 bp nucleotides: 16,764,933 to 16,770,147 of chromosome 1). We amplified two chimeric NBPF1 transcripts in the 32-7A cell line ( Figure 3). In one chimeric transcript, the type-7 exon sequence was spliced to chromosome-17 sequences located 338 bp distal to the breakpoint, yielding a transcript extended by 295 nucleotides derived from chromosome 17, and an open reading frame extended by 34 additional codons (GenBank Acc No. AF420438). In a second chimeric transcript, the type-7 exon sequence was spliced to chromosome-17 sequences located 8,602 bp from the breakpoint, yielding a transcript extended by 325 nucleotides derived from chromosome 17, and an NBPF1 open reading frame extended by 11 additional codons (GenBank Acc Nu AF420439). The sequences derived from chromosome-17 were not found in other sequences known to be expressed, and were therefore classified as non-expressed sequences. An in silico analysis of the chimeric proteins encoded by these transcripts did not reveal any other identifiable domains.

Chimeric transcripts resulting from the translocation
For ACCN1, we were unable to detect any chimeric cDNAs in the two hybrid cell lines containing the derivative chromosomes.

Expression profiling in neuroblastoma cell lines and tumors
A typical feature of a classical TSG is that its function is inactivated in different tumors. This inactivation can be caused by lack of expression due to deletion or promoter silencing. In a first quantitative RT-PCR assay we assessed four amplicons distributed across the NBPF transcripts ( Figure 4A, top panel). This allowed us to define the expression profile of the complete NBPF gene family in neuroblastoma cell lines, as the primers we used bind to regions that show high sequence identity between the different NBPF paralogs. Additionally, due to the repetitive nature of the NBPF genes [27], some primers can also bind to multiple positions on one transcript. We chose two amplicons that occur only once in a single NBPF transcript: on exon types 8-9 (amplicon B) and on the 39 UTR (amplicon D). In addition, we used two amplicons that can occur in multiple copies in the same transcript: one amplicon on the 59 repeat exons (amplicon A) and one amplicon on the 39 repeat exons (amplicon C). In the reference NBPF1 transcript (GenBank Acc Nu NM_017940), amplicon A is present in two copies and amplicon C as one copy ( Figure 4A). However, in transcripts from other NBPF paralogs, amplicon A can be present as only one copy and amplicon C can be present as multiple copies [27]. These amplicons were analyzed in a panel of 31 neuroblastoma cell lines, some with a heterozygous deletion of NBPF1 (n = 25) and some with two copies of the NBPF1 gene (n = 5). As expected, the expression levels of amplicons B and D were comparable (Pearson correlation coefficient 0.95), but in some cases these levels differed considerably from those of amplicons A and C (Pearson correlation coefficient 0.86 for both A-D and C-D). All together, this analysis showed that the expression of all NBPF transcripts combined varied widely between different neuroblastoma cell lines: expression of amplicon C in cell lines SK-N-FI, SK-N-SH and NLF was up to fifteen-fold higher than in STA-NB-3. However, the expression level of amplicon B in these three cell lines was only six-to eleven-fold higher than in STA-NB-3. This difference was probably due to increased expression of transcripts containing multiple copies of amplicon C (e.g. NBPF10) [27] in the SK-N-FI, SK-N-SH and NLF cell lines as compared to STA-NB-3.
For all amplicons, the median expression level of the NBPF family for cell lines with NBPF1 loss was lower than for cell lines with a normal 1p chromosome ( Figure 4B). For amplicons A, B and D, the difference was statistically significant (p,0.05) but for amplicon C it was not (p = 0.13). Moreover, meta-analysis using the Oncomine database [32] revealed a significantly lower expression level of NBPF transcripts in neuroblastomas that relapsed within five years than in those that did not relapse ( Figure 4C) [33].
The abovementioned primers amplified all NBPF transcripts, but the constitutional translocation disrupted only the NBPF1 Figure 3. Genomic overview of the translocation breakpoints of the der(1) and der (17) chromosomes. Schematic representation of the normal chromosome 1p36 and 17q12 regions is shown in the central box. On chromosome 1 (black bars) both translocation breakpoints are located in intronic regions of the NBPF1 gene, which is only partially shown (indicated by the dotted line). The breakpoint of the der(1) chromosome is located in a LINE repeat, shown as a black box labeled L1-PA4 on the genomic region. The black boxes below the genomic sequence represent the exons of the NBPF1 gene, with the exon type [27] indicated under each box. The translocation resulted in deletion of 5,215 bp. The translocation breakpoint on chromosome 17 (white bars) is located in an ACCN1 intron and disrupts isoform 2 but leaves isoform 1 intact. Transcription starts are indicated by angled arrows. The breakpoint is located in a LINE repeat, shown as a white box labelled L2 on the genomic region. White boxes below the genomic sequence represent the exons of the ACCN1 gene. On the der(1) chromosome (top), the NBPF1 gene is disrupted, giving rise to two chimeric cDNAs (shown as NBPF1-CHI, with the grey box representing the chromosome 17-derived fusion partners). The promoter region and first exon of ACCN1 isoform 2 are translocated to the der(1) chromosome, but no chimeric transcripts have been isolated yet (shown with the question mark and dotted line). On the der(17) chromosome (bottom), ACCN1 isoform 1 is unaffected by the translocation, but as the promoter and first exon of isoform 2 are translocated, isoform 2 is probably no longer expressed (shown by the dotted line and question mark). The last three NBPF1 exons (shown as exons of types 11, 12, and 14) are translocated to the der(17) chromosome. Arrows on the genomic sequence represent promoter sequences, and arrowheads in the transcripts show the orientation of the genes; tel, telomeric end; cen, centromeric end. The figure is not drawn to scale. doi:10.1371/journal.pone.0002207.g003  [27]. The top panel shows the location of these amplicons in the standard NBPF1 transcript. Amplicon A is present in two copies in this particular transcript. Expression levels were normalized to four reference genes [19], and the lowest value was set to 1. (B) Neuroblastoma cell lines were grouped according to the NBPF1 status (heterozygous NBPF1 deletion or not). For all amplicons, this resulted in a gene, so we also determined the expression profile of this gene specifically. We designed primers located on the 59 UTR of the NBPF1 gene, where it has slightly lower sequence homology with the other NBPF paralogs. Specific NBPF1 amplification was determined by melting curve analysis, which showed a single peak, and by sequence analysis of the amplicon. The expression level of NBPF1, determined for a panel of 32 neuroblastoma cell lines, varied widely between the different samples ( Figure 5), like it did in the expression analysis of the total NBPF family. Interestingly, we did not observe complete loss of NBPF1 expression in any cell line, as has been described for other genes located in the 1p36 region [16,17]. Statistical analysis of these data with the Wilcoxon rank sum test revealed a significantly lower expression level of NBPF1 in cell lines with a heterozygous deletion of the NBPF1 locus compared to cell lines without NBPF1 loss (p,0.05).
Our analysis of ACCN1 ( Figure 6) was limited to meta-analysis using the Oncomine database [32]. We observed a significantly weaker expression of ACCN1 in neuroblastomas with MYCN amplification (p,0.01) [34] or with 1p36 loss of heterozygosity (p,0.01) [34] compared to tumors without these alterations. Additionally, weaker ACCN1 expression was correlated with higher stages of disease (p,0.01) [33]. On the other hand, no difference was observed between neuroblastomas with or without 11q23 loss of heterozygosity (LOH) (p = 0.10) [34].

NBPF1 expression suppresses anchorage-independent growth
The DLD1Tr21 cell line is a human colon cancer cell line expressing high levels of the tetracycline repressor protein, resulting in a tight control of the expression of the gene of interest under control of a responsive tetracycline operator element [21]. This reliable cell system was stably transfected with an expression vector harboring an inducible flag-tagged full length NBPF1. As shown in Figure 7A, the addition of doxycyclin (Dox) to the cell cultures for 48 h resulted in the expression of NBPF1 in the cytoplasm of the cells.
To determine whether overexpression of NBPF1 influences anchorage-independent growth, we compared the ability of DLD1Tr21/mock and DLD1Tr21/NBPF1 cells to form colonies in soft agar in the absence and presence of Dox. NBPF1expressing cells formed significantly fewer colonies than the control cell lines or the non-induced DLD1Tr21/NBPF1 (figures 7B and 7C). These data indicate that overexpression of NBPF1 strongly inhibits the ability of this human cancer cell line to form colonies in soft agar. Figure 5. Expression analysis of NBPF1 in a panel of 32 neuroblastoma cell lines. Expression levels were normalized to two reference genes [19], and the lowest value was set to 1. Inset: relative expression levels of NBPF1 in neuroblastoma cell lines with normal NBPF1 status (grey; NO) or with a heterozygous deletion of the NBPF1 locus (white; YES). Significantly lower NBPF1 expression is observed in cell lines with NBPF1 loss (p,0.05). doi:10.1371/journal.pone.0002207.g005 median expression level that was lower in cell lines with NBPF1 loss, and this difference was significant for amplicons A, B and D (p,0.05), but not for amplicon C (p = 0.13). (C) Meta-analysis of the NBPF expression level in neuroblastoma tumors using the Oncomine database [32]. The reporter used in the array experiment was derived from NBPF1, but due to the high sequence identity among the NBPF paralogs, this analysis also represents the expression level of all NBPF genes. Significantly lower levels of NBPF transcripts are observed in tumors that relapse within five years. doi:10.1371/journal.pone.0002207.g004

Discussion
Constitutional chromosomal abnormalities in patients bearing specific tumors have often been used to clone new TSGs. Here, we describe the breakpoint cloning of a constitutional translocation t(1;17)(p36.2;q11.2) present in a neuroblastoma patient. Aberrations in the chromosomal regions involved in this translocation are recurrent in neuroblastoma, and so we hypothesized that this translocation predisposed the patient to the development of neuroblastoma. We cloned both translocation breakpoints and identified on chromosome 1 a novel gene that is interrupted by the translocation. This gene is a member of the recently described NBPF gene family, which comprises 22 genes with highly repetitive exon types [27][28][29]. As these genes are located in regions of segmental duplications on chromosome 1, they show high sequence homology in both coding and non-coding regions.
Unbalanced translocations between chromosomes 1 and 17 are common in neuroblastomas, but currently it is unclear whether there is a link between these somatic translocations and the constitutional balanced translocation described here. Only a few of the translocation breakpoints have been cloned [35,36], and no unifying underlying mechanisms for the translocations have been proposed. Two groups reported increased frequency of chromosome 11q translocations in regions of segmental duplications [37,38]. In this respect, is striking that the chromosome 1 breakpoint in the t(1;17) translocation described here is also localized in a region of segmental duplication.
NBPF1 is located on human chromosome 1p36. Though it is located outside of the smallest region of overlap defined by a largescale analysis of neuroblastoma tumors [15], at least one copy of the region encompassing the NBPF1 gene is deleted in most neuroblastomas with 1p deletion. Tumor material from the neuroblastoma patient with the constitutional translocation was not available for genetic analysis, so we cannot exclude the possibility that the der(17) was lost in the tumor, leading to loss of the 1p segment distal to the translocation breakpoint. All genes located distal to the NBPF1 gene would therefore show LOH in the tumor, including a number of recently identified TSGs located in this region [13,14]. Alternatively, as our NBPF1 expression profiling revealed a decreased expression in neuroblastoma cell lines with a heterozygous deletion of the NBPF1 locus, it is also possible that NBPF1 acts as a haploinsufficient gene. The expression pattern of NBPF1 is reminiscent of the expression analysis of other genes located in the 1p35-36 region, where 15-20% of the genes showed a lower expression level in neuroblastoma tumors and cell lines with 1p loss of heterozygosity [17,39]. It has been proposed that the combined diminished expression of these genes, rather than the inactivation of one single classical TSG, could cause the unfavorable outcome associated with 1p deletions in neuroblastoma. In addition to analyzing specifically the NBPF1 gene, we also analyzed the complete NBPF gene family in neuroblastoma cell lines; most amplicons showed a significantly lower expression level in neuroblastoma cell lines with NBPF1 loss. In these cell lines the deletion was limited to NBPF1 and NBPF3, and the other NBPF members, located on chromosome 1q21, apparently cannot fully compensate for the decreased expression of the two deleted NBPF genes. Our qRT-PCR analyses were complemented by a meta-analysis using the Oncomine database [32]. These analyses showed decreased NBPF levels in neuroblastomas that relapsed within five years. During our previously reported analysis of the NBPF family, we detected structural variation in the NBPF1 gene in different individuals [27]. Similar results were reported in large-scale studies investigating copynumber variation between people [30]. These structural variants of NBPF1 and its paralogs should be taken into consideration when choosing amplicons and analyzing the expression levels of these genes in different cell lines.
Furthermore, expression of NBPF1 in a human colorectal cell line severely suppressed soft agar colony formation, demonstrating that NBPF1 might act as a TSG, at least in colon cancer. So far, technical difficulties have prevented us from testing this hypothesis in neuroblastoma cell lines. As colorectal cancer is also characterized by frequent deletions or translocations of 1p36 [11], we believe that our present data is a first step in the elucidation of the potential tumor suppressive properties of NBPF1. The underlying mechanism for this NBPF1 activity is currently unknown.
ACCN1 encodes an amiloride-sensitive cation channel and can be expressed as two different isoforms differing in their aminoterminal domains [26]. The constitutional translocation in the present patient truncates the longer isoform. Interestingly, restoration of surface expression of this isoform in glioma cells reduces cell growth and migration [40], classifying it as a potential tumor suppressor. Meta-analysis of ACCN1 expression in neuro-blastomas showed a correlation between 1p36 LOH and decreased levels of ACCN1, but no such correlation was observed between 11q23 LOH and the expression level of ACCN1 [34]. LOH of 1p36 and of 11q23 are both associated with high-risk disease, but only 1p36 LOH is associated with MYCN amplification [6]. Another report showed a decreased expression level of ACCN1 in neuroblastomas with high telomerase activity [41]. Knockout mice defective in both ACCN1 isoforms were viable, fertile, and showed no morphological or behavioral abnormalities, but did indicate that ACCN1 has a role in retinal function [42]. No data have been reported on tumor incidence or tumor aggressiveness in these mice. The most frequent chromosomal abnormality in neuroblastomas is gain of chromosome 17 [43], which appears to conflict with a potential tumor suppressor function for ACCN1 in neuroblastoma. However, most 17q breakpoints are distal to ACCN1, and it has been suggested that the 17q12 region might contain a gene that suppresses neuroblastoma [44]. These data warrant further investigation into the role of ACCN1 in neuroblastoma. Additionally, it remains possible that the translocation analyzed here affects an unidentified gene in the ACCN1 intron in which the translocation breakpoint lies. In this region we observed two putative promoter sequences, as predicted by FirstEF [45], and a small number of ESTs not annotated to a known gene and devoid of a meaningful open reading frame (data not shown). It is not clear whether these sequences have a function or are merely the result of random transcriptional activation.
In conclusion, we describe the molecular cloning of both translocation breakpoints in a neuroblastoma patient with a constitutional t(1;17) translocation. Genes overlapping both translocation breakpoints were identified, namely NBPF1 on chromosome 1 and ACCN1 on chromosome 17. Further research might reveal the role of the disruption of both of these genes in the development and progression of neuroblastoma.