ARID3B Induces Malignant Transformation of Mouse Embryonic Fibroblasts and Is Strongly Associated with Malignant Neuroblastoma

ARID3B, a member of the AT-rich interaction domain (ARID) family of proteins, plays an essential role in the survival of neural crest during embryogenesis. Here, we report evidence that ARID3B is involved in the development of malignant neuroblastoma, a childhood tumor derived from neural crest. (a) ARID3B is expressed by all five cell lines derived from neuroblastoma tested by us. (b) Analysis of published DNA microarray data of fresh neuroblastoma tumors showed that ARID3B is expressed in 80% of stage IV tumors, whereas only in 9% of stage I-III+IVs tumors. (c) In vitro growth of several neuroblastoma cell lines is suppressed significantly by antisense as well as siRNA treatment. (d) An increase of the ARID3B expression level by transfection in the SY5Y neuroblastoma cell line enhances the malignancy in tumor growth assays in nu/nu mice. (e) ARID3B by itself can immortalize mouse embryonic fibroblasts (MEFs) in vitro and confers malignancy to MEF when transfected together with MYCN, the best characterized oncogene for neuroblastoma. Thus, ARID3B seems to play a key role in the malignant transformation of neuroblastoma and may serve not only as a marker of malignancy but also as a potential target for cancer therapy of stage IV neuroblastoma for which there is currently no effective treatment available. (Cancer Res 2006; 66(17): 8331-6)

Neuroblastoma is the most common malignant infant tumor (1). It affects the sympathetic nervous system derived from neural crest and has a remarkable variability in its clinical course (1, 2). Roughly half are discovered as solitary tumors without metastasis, of which a considerable proportion undergo spontaneous regression. In contrast, the other half are discovered with metastasis and take a poor clinical course. Although the MYCN gene has been implicated in the development of neuroblastoma, the molecular mechanisms underlying such a diversity remain unknown (1, 3).

Recently, we have reported that neural crest development is severely suppressed in embryos that bear null mutations in the ARID3B gene (4). This phenotype suggests that the role of ARID3B is to maintain the survival of proliferating neural crest. Interestingly, the most similar ARID protein, ARID3A was shown to induce malignant transformation of mouse embryonic fibroblasts (MEF) in collaboration with activated oncogenes such as Ras (5). These two observations inspired us to examine the possibility that ARID3B is involved in the formation of neuroblastoma, which is the archetypal neural crest tumor. Here, we provide strong evidence that ARID3B is involved in development of malignant neuroblastoma.

Cell culture, gene transfection, and tumor growth assay. Human neuroblastoma cells (SH-SY5Y, TGW, CHP-126, NBLS, and IMR-32) were maintained in a 1:1 mixture of DMEM and Ham's F-12 medium (Invitrogen, Carlsbad, CA) supplemented with 10% FCS. Human hematopoietic cells, K-562 (American Type Culture Collection, Manassas, VA), were maintained in RPMI 1640 (Invitrogen) supplemented with 10% FCS. Primary MEFs were isolated from 14.5 dpc mouse embryos and cultured in DMEM supplemented with 10% FCS. Cells were maintained in a humidified 5% CO₂ atmosphere.

Soft agar assays were done as previously described (6). The retroviral expression vectors, MSCV-IRES-EGFP and MSCV-IRES-DsREDT4, were used for transfecting ARID3B fused with the flag epitope and MYCN fused with the hemagglutinin-epitope, respectively (7). Production and infection of retroviruses were done as described previously (8). Following purification of GFP⁺ or DsREDT4⁺ cells by fluorescence-activated cell sorting (FACS), the expression of ARID3B and MYCN was confirmed by Western blot analysis with antibodies against each tag. Primary MEF (passage 2) were infected with a retrovirus carrying ARID3B or MYCN. In some experiments, MEF that forcibly expressed ARID3B were further infected with MYCN-carrying virus at passage 18. For evaluating tumorigenicity, either 10⁶ MEFs or 10⁷ neuroblastoma cells in 100 μL PBS were s.c. transplanted into BALB/c-nu/nu mice. Tumor growth was monitored every 3 days.

RNA quantification. Northern blot analysis was done as described previously using a part of the hARID3B cDNA (positioned from 1,298 to 3,480 bp, Genbank accession no. BC041792) as the probe (4). Primers used for reverse transcription-PCR (RT-PCR) and quantitative PCR analyses are listed in Supplementary Fig. S1A and B.

Antisense oligonucleotides and short interfering RNA. The ARID3B antisense (AS) oligonucleotides used for transfection, were designed and synthesized by Gene Design, Inc. (Osaka, Japan), and contained 2′-O,4′-C-methylene bridged nucleic acid (9). AS1 (5′-CCATTTTTgcCTCaagcttC-3′) and AS2 (5′-TGCccAgGTTcagagtCATG-3′) were targeted at starting locations 141 and 108 of human ARID3B mRNA (Genbank accession no. BC041792), respectively. Scrambled oligonucleotides (5′-gCTTCcCtTcTtgCAaCaTT-3′ to AS1 and 5′-GccACaTgGGcTggTaCAtT-3′ to AS2) have the same base composition as hARID3B AS, but in a random order. Capital letters indicate bridged nucleic acid residues.

Short interfering RNA (siRNA) sequences targeting ARID3B correspond to the coding regions 944 to 964 (siRNA1) and 1,037 to 1,057 (siRNA2) relative to the first nucleotide of the start codon. Scrambled siRNAs (5′-GAGCGAUGCCUAGAAUAAU-3′ to siRNA1 and 5′-GCUCGCAAUAUGUAACAGU-3′ to siRNA2) were designed and used as a control. All sequences were compared with the human genome (using Basic Local Alignment Search Tool) to ensure their specificities. Antisense oligonucleotides and siRNAs were transfected using LipofectAMINE 2000 (Invitrogen) according to the protocol of the manufacturer. In AS experiments, the relative growth ratio against scramble-treated cells was calculated by the ATP assay, using the CellTiter-Glo Luminescent Cell Viability Assay kit (Promega, Madison, WI). Each assay consists of triplicate cultures. In some experiments, the proportion of apoptotic and dead cells was measured by flowcytometer using Annexin, 7-amino-actinomycin D, and propidium iodide stainings.

Analysis of DNA chip data. Expression data from McArdle et al. (10) was obtained from ArrayExpress,³

accession code E-MEXP-83. Expression values, detection calls, and detection P values were extracted from the entire data set using a series of perl scripts. Additional data from human tissues and tumor samples was also obtained from ArrayExpress (accession nos. E-MEXP-66, E-MEXP-97, E-MEXP-101, E-MEXP-114, E-MEXP-120, E-MEXP-214, E-MEXP-167, E-MEXP-353, E-MEXP-149, E-MIMR-17, and E-AFMX-5). The data sets were selected to cover a wide range of different tumors and normal tissues and to avoid samples that had been subjected to experimental treatments. A full list of samples is available in the Supplementary Data. The association between different types of neuroblastoma and the expression of various genes was calculated using the hypergeometric distribution. This calculates the probability of observing an overlap of a given size or larger between two groups of a given population. For the purpose of statistical calculations, a call of marginal was considered as detected (i.e., present). The neuroblastoma cell lines included in McArdle et al. (10) were considered as stage IV neuroblastoma when association P values were calculated for all probe sets (Fig. 1D) but were excluded from calculations for specific P values quoted in the text.

To test ARID3B expression in neuroblastoma, we obtained five human neuroblastoma cell lines and examined the expression of ARID3A, ARIDB, and MYCN. All obtained cell lines express significant although variable levels of ARID3B (Fig. 1A). In contrast, ARID3A was not detected in any of the cell lines even by RT-PCR (Fig. 1B). Of the five cell lines, strong MYCN expression was detected in four cell lines, whereas one line expresses a low level that is detectable by RT-PCR. In contrast, we detected ARID3B expression in only two (K-562 and SW480) of eight nonneuroblastoma cancer cell lines (Supplementary Fig. S1C).

We next examined the expression of ARID3B, ARID3A, and MYCN in clinical samples of neuroblastoma using published microarray data. We used the data set of McArdle et al. (10) as these data is available from ArrayExpress (11),³ has been described in detail and uses the highly standardized Affymetrix oligonucleotide arrays, enabling us to extract the relevant data and to compare this across other data sets. Of 23 malignant neuroblastoma samples, ARID3B was detected as present (++) or marginal (+) in 11 samples (Fig. 1C). Of these, two are neuroblastoma cell lines (not shown in Fig. 1C), one is a stage II neuroblastoma, and the remaining eight represent stage IV neuroblastomas. Thus, ARID3B expression was found in 9 of 21 (not counting the cell lines) neuroblastoma cases (42.8%; Fig. 1C). Strikingly, this expression is strongly associated with stage IV tumors (8 of 10 of stage IV versus 1 of 11 of stage I-II+IVs, P = 0.0018) and also with tumors with 11q deletion (9 of 11, P = 0.00019). ARID3A is expressed in five of the ARID3B⁺ samples but in none of the ARID3B⁻ tumors, suggesting a possible relationship (P = 0.06). MYCN was detected in the two cell lines and in 19 of the 21 neuroblastomas, including in all ARID3B⁺ tumors.

In mouse, the expression of ARID3B is detected in only a small number of embryonic and adult tissues (4). To determine whether the same is true in human, we obtained microarray data created using the same array used by McArdle et al. (10) from a total of 12 different experiments covering a wide range of normal tissues and cell types (∼70) and tumors (∼30) represented by a total of 483 samples (Supplementary Data). Of these 483 samples, ARID3B was detected as expressed (present or marginal) in only 25 samples. Eleven of these are the previously described neuroblastoma samples. Hence, ARID3B can be said to be specific to stage IV neuroblastoma, both in the narrow sense of not being expressed in earlier stage neuroblastoma, but also in the broader sense of being expressed in only a few other cell types. If there are only a few molecules that are specific to neuroblastoma in this manner, then this strengthens our contention that ARID3B is a key molecule in the progression of neuroblastoma. To ascertain this, we calculated P values (Supplementary Data) for all probe sets for the specific association with stage IV neuroblastoma and a general association (compared against the data for the 30 tumor types in our data sets). Although there are five probe sets that are more strongly associated with stage IV versus other stages of neuroblastoma than ARID3B, these are also expressed in many other tumor types. Surprisingly, ARID3B is easily distinguished from all other probe sets as being the only gene that is specific to stage IV neuroblastoma in both a specific and a broader sense (Fig. 1D).

To investigate the role of ARID3B in the growth of neuroblastoma cell lines, we did suppression and overexpression of ARID3B. First, we suppressed ARID3B expression using two types of siRNAs (Fig. 2A-D). Although the SY5Y cell line was insensitive to this treatment, siRNA treatment suppressed the growth of other neuroblastoma cell lines to a variable extent (85-10%; Fig. 2A and C). We also treated these cell lines with AS molecules to ARID3B and obtained similar results to the siRNA experiments (Supplementary Fig. S1D). These results clearly indicate the involvement of ARID3B in the growth of some, although perhaps not all, neuroblastoma cell lines. Interestingly, expression of additional ARID3B by transfection showed no significant effect on the cell lines whose growth was strongly affected by siRNA or AS (Supplementary Fig. S1D). However, both anchorage-dependent and anchorage-independent growth of the SY5Y cell line was markedly enhanced by transfection of the ARID3B gene (Fig. 3A and B; Supplementary Fig. S2A). Moreover, malignancy, as assessed by the mean survival times of tumor transplanted nu/nu mice, was accelerated from 47 to 33 days (Fig. 3C). Taken together, these data suggest that ARID3B is a key molecule supporting the growth of neuroblastoma.

Finally, we analyzed the transforming activity of ARID3B by exploiting the MEF transformation assay (6). Primary MEF cells were transfected with a retrovirus vector containing ARID3B (Supplementary Fig. S2B). In the control groups (untransfected or vector alone), the proliferative ability of MEF decreased over passage (Fig. 4A), whereas transfection by ARID3B immortalized MEF (Fig. 4A). Thus, ARID3B by itself can transform MEF, which is similar to the activity of ARID3A (5). We next assessed whether ARID3B can collaborate with MYCN to alter the growth of MEF. Additional transfection of the MYCN gene to ARID3B⁺ MEF indeed enhanced their growth (Fig. 4B). More importantly, coexpression of ARID3B and MYCN in MEF conferred both anchorage-independent growth (Fig. 4C) and tumorigenic ability in nu/nu mice (Fig. 4D; Supplementary Fig. S2C). Thus, ARID3B is a transforming gene and is likely to be involved in the malignant transformation of neuroblastoma.

Although ARID family molecules, such as JARID1B (12, 13) and ARID4B (12, 14), have been detected in specific types of carcinoma, there is only one report directly demonstrating the transforming activity of ARID family molecules (5). Peeper et al. (5) showed that expression of ARID3A by itself immortalizes MEF and induces malignant transformation when expressed together with oncogenic RAS. Despite such a clear oncogenic activity in vitro, its involvement in actual tumors has not been documented. Our present study shows that ARID3B satisfies the two main criteria for tumorigenic molecules, the ability to transform MEF and the specific association with the given cancer. The fact that ARID3B collaborates with MYCN to induce malignant transformation of MEF and that it is expressed preferentially in stage IV tumors is consistent with the idea that ARID3B plays a key role in the development of malignant neuroblastoma. Our results suggest that MYCN expression is required for the tumorigenic activity of ARID3B. Although our data analysis suggests that ARID3B expression is correlated with 11q deletion rather than with MYCN amplification, there is no inconsistency with our experimental findings since all ARID3B⁺ tumors also express MYCN. Indeed, it is actually difficult to establish whether ARID3B expression is associated with 11q deletion or advanced-stage neuroblastoma because almost all (8 of 10) stage IV samples have 11q deletions. However, because two of three non–stage IV tumors carrying 11q deletions do not express ARID3B, ARID3B expression may be more coupled to tumor progression than 11q deletion. Interestingly, ARID3A, whose AT-rich interaction domain has 84% homology with that of ARID3B, is also able to collaborate with MYCC to alter the proliferation of MEF (4, 5). However, the combination of MYCC and ARID3A cannot confer the ability to undergo anchorage-independent proliferation, nor to form tumors in nu/nu mice (5). It is thus likely that ARID3A is less tumorigenic in combination with MYC-induced transformation than ARID3B.

Our data strongly implicates dysregulated ARID3B expression in the development of malignant neuroblastoma, but addresses neither the function of ARID3B in the development of neuroblastoma nor the causes underlying its expression. However, previous findings suggest that ARID3B is involved in avoiding oncogene-induced apoptosis: (a) ARID3A, the closest relative of ARID3B, has been shown to be able to bypass Ras or MYCC-induced apoptosis (5); (b) AS treatment of the CHP126 cell line resulted in accelerated apoptosis (Supplementary Fig. S2D); and (c) the function of ARID3B during embryogenesis is to protect neural crest cells from apoptosis at their proliferative phase (4). As ARID3B was initially isolated as a RB1-binding protein (15), it seems likely that the Rb pathway is involved in apoptotic avoidance in neuroblastoma.

In normal mouse, the ARID3B gene is expressed under strict spatiotemporal control, with expression limited to nascent mesoderm and neural crest (4). Our data suggests that ARID3B expression in neuroblastoma is aberrant, as it is expressed only in a specific type or stage of neuroblastoma. Thus, the trigger for the aberrant expression of ARID3B seems likely to be an important event in the development of malignant neuroblastoma.

Although ARID3B has not been recognized by previous extensive searches for molecules involved in neuroblastoma formation (10, 16–18), our analysis of previously published data suggests that it might have been possible to identify ARID3B from existing data if different questions had been asked. Our experience suggests that it would have been necessary to analyze a range of data sets from a variety of sources to identify ARID3B as involved in neuroblastoma from a pure data mining exercise. The ability to combine data sets from a wide range of sources is a major advantage of standardized array analyses that is often overlooked but has great potential for future research.

In conclusion, this study shows that ARID3B is a tumorigenic molecule that can collaborate with MYCN to induce malignancy and implies its role in the progression of malignant neuroblastoma. The strong association of ARID3B expression with stage IV neuroblastoma suggests that ARID3B may serve as a good tumor marker for neuroblastoma. As siRNA and AS treatments can suppress growth of some neuroblastoma cell lines, ARID3B may also serve as a target for the therapy of stage IV neuroblastoma for which development of effective treatments have been lacking.

Grant support: Ministry of Education and Science grants 16606005 and 17045039 (T. Era), Knowledge Cluster Initiative grant (T. Era), and Leading Project for Realization of Regenerative Medicine grant (S.I. Nishikawa).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Dr. H. Enomoto and Dr. M. Royle for a critical reading of the manuscript and a helpful discussion.