Characterization of a Germ-Line Deletion, Including the Entire INK4/ARF Locus, in a Melanoma-Neural System Tumor Family: Identification of ANRIL, an Antisense Noncoding RNA Whose Expression Coclusters with ARF

We have previously detected a large germ-line deletion, which included the entire p15/CDKN2B-p16/CDKN2A-p14/ARF gene cluster, in the largest melanoma-neural system tumor (NST) syndrome family known to date by means of heterozygosity mapping based on microsatellite markers. Here, we used gene dose mapping with sequence-tagged site real-time PCR to locate the deletion end points, which were then precisely characterized by means of long-range PCR and nucleotide sequencing. The deletion was exactly 403,231 bp long and included the entire p15/CDKN2B, p16/CDKN2A, and p14/ARF genes. We then developed a simple and rapid assay to detect the junction fragment and to serve as a direct predictive DNA test for this large French family. We identified a new large antisense noncoding RNA (named ANRIL) within the 403-kb germ-line deletion, with a first exon located in the promoter of the p14/ARF gene and overlapping the two exons of p15/CDKN2B. Expression of ANRIL mainly coclustered with p14/ARF both in physiologic (various normal human tissues) and in pathologic conditions (human breast tumors). This study points to the existence of a new gene within the p15/CDKN2B-p16/CDKN2A-p14/ARF locus putatively involved in melanoma-NST syndrome families and in melanoma-prone families with no identified p16/CDKN2A mutations as well as in somatic tumors. [Cancer Res 2007;67(8):3963–9]

Hereditary cutaneous malignant melanoma (CMM) is a well-established syndrome (1). One of the main CMM loci has been mapped to chromosome arm 9p21. Three candidate tumor suppressor genes have been identified in this region: p15/CDKN2B encodes p15 protein, p16/CDKN2A encodes p16 protein, and p14/ARF encodes p14^ARF protein (p19^ARF in mice; ref. 2). p14^ARF protein is encoded by an alternative exon 1 (1β) and by exons 2 and 3 of the p16/CDKN2A gene in a different reading frame. Consequently, p16 and p14^ARF share no homology at the amino acid level and have significantly different functions. p15 and p16 are specific cyclin-dependent kinase (CDK) 4/CDK6 inhibitors in the retinoblastoma (Rb) pathway, and inactivation of p15 and/or p16 allows cells to escape cell cycle arrest in G₁. p14^ARF acts in both the p53 and Rb pathways by interacting with MDM2, a protein common to the two pathways. p14^ARF binds MDM2 and promotes its degradation, resulting in p53 activation and G₁ and G₂ arrest. p16/CDKN2A is currently the most clinically relevant melanoma susceptibility gene in 9p21. By contrast, germ-line mutations in p14/ARF are rare (3, 4) and have not been described in p15/CDKN2B.

Following the observation that patients with neural system tumors (NST) are at an increased risk of developing CMM, a large population-based survey showed an increased risk of NSTs in melanoma patients and also in their first- and second-degree relatives (5). This has been termed the melanoma-astrocytoma syndrome, owing to the presence of astrocytomas in the first family to be characterized (OMIM 155755; ref. 6). The authors suggested joint predisposition segregating as a Mendelian disorder. Three large deletions have now been reported in families with a hereditary predisposition to melanoma and NSTs: one involving p15/CDKN2B, p16/CDKN2A, and p14/ARF in a French family (7), one 21,826-pb deletion restricted to exon 1β of p14/ARF in a U.S. family (ref. 8, family B in ref. 9), and one probably involving only p14/ARF in a U.K. family (ref. 7, family D in ref. 9). Three additional large deletions have been reported in families with a hereditary predisposition to melanoma without NSTs: two involving only p14/ARF in a U.S. (10) and in a U.K. family (family A in ref. 9) and one restricted to p16/CDKN2A in a U.K. family (family C in ref. 9). It is noteworthy that Mistry et al. (9) suggested coancestry of families A (without NSTs) and B (with NSTs) and concluded that the susceptibility in these two families was primarily to melanoma. Finally, germ-line mutations in p16/CDKN2 have been reported in families with a hereditary predisposition to melanoma with NSTs (9).

Here, using sequence-tagged site (STS) real-time PCR-based gene dose mapping, we determined the precise size and end points of a large germ-line deletion removing the entire p15/CDKN2B-p16/CDKN2A-p14/ARF gene cluster initially detected in a large French melanoma-NST syndrome family (7).

We also identified, within the 403-kb germ-line deletion, a new large antisense noncoding RNA (named ANRIL) with a first exon located in the promoter of the p14/ARF gene and overlapping the two exons of p15/CDKN2B. Expression of this noncoding RNA coclustered mainly with p14/ARF in both physiologic and pathologic conditions.

Patients. The family with a cluster of CMM and NSTs (Fig. 1) has been investigated since 1977 at Hôpital Pitié-Salpêtrière in Paris, France. About 100 individuals have been surveyed. Pathologic materials initially investigated in 16 different laboratories were, when possible, reassessed by both a dermatopathologist and a neuropathologist. Extensive historical and clinical descriptions of this kindred are available elsewhere (11). Overall, eight members of this family have or had CMM and nine members have or had NSTs, ranging from astrocytoma of all grades to schwannoma, neurofibroma, and meningioma. Six relatives have both skin tumors and NSTs, resulting in statistical cosegregation of the skin tumor and NST traits. This family therefore illustrates a melanoma-NST syndrome probably inherited as a single autosomal dominant Mendelian trait. However, the possibility of other explanations cannot be ruled out for the constellation of tumors observed in this family, such as multiple tightly linked genes or a contiguous gene syndrome.

DNA extraction. High-molecular-weight DNA was prepared by standard proteinase K digestion followed by phenol-chloroform extraction from whole blood leukocytes or lymphoblastoid cell lines from seven individuals, including three affected haplotype-positive relatives (III-16, III-23, and IV-20) and four healthy haplotype-negative relatives (II-9, III-10, III-14, and IV-9). The haplotypes have been defined elsewhere (11).

STS real-time PCR-based gene dose mapping. In this method, quantitative values are obtained from the threshold cycle number at which the increase in the signal associated with exponential growth of PCR products begins to be detected by PE Biosystems analysis software (Perkin-Elmer Applied Biosystems, Foster City, CA), used as recommended in the manufacturer's manuals. The precise amount of genomic DNA added to each reaction mix (based on absorbance) and its quality (i.e., lack of extensive degradation) are both difficult to assess. We therefore also quantified the ALB gene (encoding albumin and mapping to chromosome region 4q11-q13) as an endogenous DNA control, and each sample was normalized based on its ALB content. The relative STS marker copy number was also normalized to a calibrator, or 1× sample, consisting of genomic DNA from a normal subject.

Final results, expressed as N-fold differences in the target STS marker copy number relative to the ALB gene and the calibrator, and termed “N_STS,” were determined as follows: N_STS = 2^{(ΔCtsample − ΔCtcalibrator)}, where ΔCt values of the sample and calibrator are determined by subtracting the average Ct value of the target STS marker from the average Ct value of the ALB gene. Given the target STS marker, samples with N_STS values of 0.5 and 1.0 were considered deleted and normal, respectively. Primers for ALB and the 26 9p21-linked STS markers were chosen with the assistance of the computer program Oligo 4.0 (National Biosciences, Plymouth, MN). We conducted BLASTN searches against nr (the nonredundant set of the Genbank, European Molecular Biology Laboratory, and DNA Data Bank of Japan database sequences) to confirm the total STS specificity of the nucleotide sequences chosen for the primers. The primer pairs for each STS were selected to be unique in the human genome; in particular, they were located outside repetitive DNA sequences (Alus, LINEs, etc.). All PCRs were done with an ABI Prism 7700 Sequence Detection System (Perkin-Elmer Applied Biosystems) and the SYBR Green PCR Core Reagents kit (Perkin-Elmer Applied Biosystems). The thermal cycling conditions comprised an initial denaturation step at 95°C for 10 min and 50 cycles at 95°C for 15 s and 65°C for 1 min. Experiments were done with triplicates for each data point.

Fine characterization of the 403-kb germ-line deletion. Long-range PCR was done with the Expand Long Template PCR System (Boehringer Mannheim, Mannheim, Germany) as recommended by the manufacturer. Forward primer STS20 and reverse primer STS17 amplified an ∼1.5-kb mutant genomic fragment, the normal fragment being too large (405 kb) to be amplified. PCR products were sequenced with the Applied Biosystems Prism Dye Terminator Cycle Sequencing Ready Reaction kit (Perkin-Elmer Applied Biosystems) and an Applied Biosystems DNA Sequencer Model 310 (Perkin-Elmer Applied Biosystems).

Real-time reverse transcription-PCR. The theoretical and practical aspects of real-time quantitative reverse transcription-PCR (RT-PCR) using the ABI Prism 7700 Sequence Detection System have been described in detail elsewhere (12). Briefly, results, expressed as N-fold differences in target gene (p15/CDKN2B, p16/CDKN2A, p14/ARF, or ANRIL gene) expression relative to an endogenous RNA control (PPIA gene), termed Ntarget, are determined by the following formula: Ntarget = 2^ΔCtsample, where the ΔCt value of the sample is determined by subtracting the Ct value of the target gene from the Ct value of the PPIA gene. The Ntarget values of the samples are subsequently normalized such that the Ntarget value of the normal human tissue (or of the breast tumor), which contains the smallest amount of target gene mRNA, is 1. The nucleotide sequences of the primers for PPIA and for the four target genes are shown in Table 1. To avoid amplification of contaminating genomic DNA, one of the two primers was placed at the junction between two exons. The thermal cycling conditions comprised an initial denaturation step at 95°C for 10 min and 50 cycles at 95°C for 15 s and 65°C for 1 min.

In a French family with a history of melanoma and NSTs, we have previously identified a large germ-line deletion, including the entire p15/CDKN2B-p16/CDKN2A-p14/ARF gene cluster (7). This family is the largest known melanoma-NST syndrome family, eight members having CMM and nine members having NSTs (six members have both skin tumors and NSTs), ranging from astrocytoma of all grades to schwannoma, neurofibroma, and meningioma (Fig. 1). Microsatellite polymorphism analysis showed heterozygosity at both the centromeric marker D9S976 and the telomeric marker D9S736, suggesting that the deleted region was smaller than 700 kb (Fig. 2). It is noteworthy that a fourth gene, MTAP, which encodes methylthioadenosine phosphorylase, is located in this 700-kb region. Behrmann et al. (13) has suggested that MTAP inactivation could be an important step in the development of melanomas.

To determine the precise size and end points of this germ-line deletion, we estimated the number of copies (one or two) by real-time PCR-based gene dosage assay for various STS located between the markers D9S976 and D9S736.

We first analyzed 11 STS markers regularly spaced (every ∼70 kb) within the deleted region. Two additional steps, using carefully chosen additional STS markers (eight and seven markers, respectively), allowed us to refine the centromeric and telomeric breakpoint regions (∼1 kb) sufficiently for long-range PCR. The strategy and results are summarized in Fig. 2. Long-range PCR was done on genomic DNA from affected patient by using the forward primer of the retained STS20 marker and the reverse primer of the retained STS17 marker, generating a PCR product of ∼1.5 kb. The normal fragment was too large (∼405 kb) to be amplified. Sequencing of this PCR product revealed that a sequence between nucleotides 22,329,545 and 21,926,315 [numbered in the Build 34 National Center for Biotechnology Information (NCBI) Map Viewer]⁴

was deleted in the PCR product (Fig. 3). Thus, the deletion was exactly 403,231 bp long (∼403 kb). There were no highly homologous sequences between the two deletion boundaries, and only a 1-bp (adenine) identical sequence at the deletion end points, making it unlikely that the deletion resulted from an unequal homologous recombination event. It is noteworthy, however, that the centromeric breakpoint is located within a truncated LINE element.

The 403-kb germ-line deletion included the entire p15/CDKN2B-p16/CDKN2A-p14/ARF gene cluster but unambiguously excluded additional well-characterized genes, namely MTAP (NM_002451) located in the telomeric region and DMRTA1 (NM_022160) in the centromeric region (Fig. 3).

To develop a direct predictive test for this large French family with the melanoma-NST syndrome associated with a 403-kb germ-line deletion, additional primers (Δ403-U: 5′-CCAGTCTATCGTTGTTGGACAT-3′ and Δ403-L: 5′-TGTATTTGATGGCATTCCACA-3′) were designed to obtain a short PCR product (210 bp) in a simple and rapid assay designed to detect the junction fragment. Our results indicate that PCR-based detection of the junction fragment is an effective and direct method for the identification of subjects with an individual risk in this family (Fig. 4). Indeed, the three affected individuals tested yielded the 210-bp fragment, whereas the four healthy relatives tested did not.

A search of the dbEST database, using the Gapped BLAST program (14), allowed us to identify several human expressed sequence tags (EST) of an unidentified transcript (Genbank accession nos. CN277071, AA807394, BF002332, BI765545, BX100299, AI825844, AW169296, and BC038540). Some of these ESTs contained a polyadenylation site, suggesting that this gene is transcribed in the direction opposite from the p15/CDKN2B-p16/CDKN2A-p14/ARF gene cluster. RT-PCR amplification of a cDNA obtained from normal human testis mRNA, using various primers placed in the identified ESTs, and sequence analysis of the PCR products, identified a large cDNA 3,834 bp long. This cDNA seems to be untranslated, as it does not possess a clear open reading frame. We chose to name this novel gene ANRIL, for “antisense noncoding RNA in the INK4 locus” (Genbank accession no. DQ485453). Alignment of the ANRIL 3,834-bp cDNA with genomic sequence led to the conclusion that the ANRIL gene contains 19 exons, with exon/intron junctions corresponding to the consensus sequences of donor/acceptor splice sites and a polyadenylation site in the last exon. The ANRIL genomic sequence spans a region of 126.3 kb. We also characterized a short alternatively spliced ANRIL mRNA species 2,659 bp long, composed of the first 12 ANRIL exons and an additional 3′ exon (located in intron 12), including a second polyadenylation site (Genbank accession no. DQ485454).

The ANRIL gene overlaps the two exons of p15/CDKN2B but does not share any nucleotide sequences (Fig. 5A). Moreover, the 5′ end of the first exon of the ANRIL gene is located about −300 bp upstream of the transcription start site of the p14/ARF gene, suggesting that these two genes may share a bidirectional promoter (Fig. 5A). Global transcriptome analysis recently provided evidence that the bulk of the mammalian genome can produce transcripts from both stands, with a large number of novel noncoding RNAs (15). The latter authors also showed the existence of a large number of transcription units (paired sense/antisense transcripts) in the mammalian genome and suggested that antisense transcription contributes to controlling transcriptional outputs (16). For example, these authors identified 1,092 coding-noncoding pairs overlapping head to head on the promoter in the mouse genome.

To test the possible coordinated transcriptional regulation of ANRIL and p14/ARF, we quantified ANRIL, p14/ARF, p16/CDKN2A, and p15/CDKN2B mRNA in 22 normal human tissues (Supplementary Fig. S1A). The highest expression of ANRIL was observed in normal ovary, and the lowest expression was observed in skeletal muscle (100-fold lower than in ovary). ANRIL coclustered with p14/ARF, p16/CDKN2A, and p15/CDKN2B in these normal human tissues, with a stronger positive correlation between ANRIL and p14/ARF (Spearman rank correlation test: r = +0.949; P < 10⁻⁷) than between ANRIL and p16/CDKN2A (r = +0.573; P = 0.004) or between ANRIL and p15/CDKN2B (r = +0.561; P = 0.005). In a series of 12 breast tumor tissues showing a wide range of p14/ARF, p16/CDKN2A, and p15/CDKN2B mRNA levels, we also observed a stronger positive correlation of ANRIL with p14/ARF (Spearman rank correlation test: r = +0.965; P = 0.000001) than with p16/CDKN2A (r = +0.902; P = 0.00008) or p15/CDKN2B (r = +0.769; P = 0.004; Supplementary Fig. S1B). Finally, in an additional series of NF1-associated tumors in tissues derived from the embryonic neural crest (the Schwann cells are considered to be the progenitors of these tumors), we confirmed the positive link between the expression of ANRIL with p14/ARF as well as with p16/CDKN2A and p15/CDKN2B (data not shown) in a second tumor type.

These data suggest coordinated transcriptional regulation of ANRIL and p14/ARF, and possibly also p16/CDKN2A and p15/CDKN2B, in both physiologic and pathologic conditions.

At the moment, it is difficult to propose a mechanism of action for noncoding RNA. Indeed, the puzzle of noncoding RNA in eukaryotes is just starting to be assembled, and certainly, further unexpected pieces will be identified in the near future. To date, it likely seems that noncoding RNAs constitute a critical hidden layer of gene regulation in complex organisms, and their understanding requires new approaches in functional assays (17). We can hypothesize that ANRIL (antisense gene) could regulate p15/CDKN2B-p16/CDKN2A-p14/ARF locus (sense genes) by various novel mechanisms, such as RNA interference, gene cosuppression, gene silencing, chromatin structure maintenance, imprinting, and DNA demethylation (18). Moreover, we cannot rule out the possibility that ANRIL gene could be able to produce mature miRNAs (19) or encode a small peptide (20). Additional studies are thus necessary to elucidate the putative role via a genetic (or epigenetic) mechanism of ANRIL gene on the regulation (deregulation in tumorigenesis) of the p15/CDKN2B-p16/CDKN2A-p14/ARF locus.

Further studies of ANRIL and/or p14/ARF and/or p15/CDKN2B will be necessary to explain the additional tumor spectrum (i.e., NSTs) in this French family. Indeed, p15/CDKN2B and p14/ARF, like p16/CDKN2A, are altered during NST progression (21–23), but no germ-line mutation of p15/CDKN2B or p14/ARF has been clearly implicated in inherited predisposition to NSTs (24). With regard to the ANRIL gene, we reanalyzed (Fig. 5B) the large 9p21 germ-line deletions previously reported in families with a hereditary predisposition to melanoma with or without NSTs (7–10). ANRIL was entirely deleted in our French family with NSTs (family F in Fig. 5B). ANRIL exon 1 was deleted in two U.K. families, one with and one without NSTs (families B and A, respectively), whereas its promoter was unambiguously deleted in one U.S. family without NSTs (families E) and probably deleted in another U.S. family with NSTs (families D). However, it could be hypothesized that these two U.S. families D and E, like the two U.K. families A and B (see ref. 9), could have an identical deletion of coancestral origin. Finally, the ANRIL gene was not deleted in a second U.K. family without NSTs (family C). Location of deletion end points of each large deletion reported in these different families needs to be fully characterized to further incriminate ANRIL as a putative candidate gene in melanoma-NST syndrome susceptibility.

The identification of this large antisense noncoding RNA could be important in cancer pathology as in hereditary predisposition to melanoma or in somatic alteration of the p15/CDKN2B-p16/CDKN2A-p14/ARF cluster observed in a large proportion of cancers, but also important in physiology area, the p15/CDKN2B-p16/CDKN2A-p14/ARF locus having a key role in the replicative senescence. Recent studies suggest that alteration of an antisense RNA can alter the expression of the corresponding sense mRNAs (25). Alteration of this noncoding RNA could be obtained by large deletion (including the entire ANRIL gene as in our family), deletion of one or several exons, deletion of a part of the promoter, splice sites mutations, aberrant DNA methylation (only in the somatic tumor), etc.

In conclusion, by applying STS real-time PCR-based gene dose mapping to a French melanoma-NST syndrome family, we identified a germ-line deletion encompassing a new putative transcriptional regulator gene (i.e., ANRIL) and the entire p15/CDKN2B-p16/CDKN2A-p14/ARF gene cluster. Our data suggest coordinated transcriptional regulation of ANRIL and p14/ARF (and, to a lesser extent, p16/CDKN2A and p15/CDKN2B) in both physiologic and pathologic conditions. Further studies are necessary to elucidate the role of the ANRIL gene in melanoma-NST syndrome families, in melanoma-prone families with no identified p16/CDKN2A mutations, and in somatic tumors.

Our new simple and rapid gene dose mapping assay should prove useful for characterizing other large germ-line deletions in research settings.

Grant support: Association pour la Recherche sur le Cancer and Ministère de l'Enseignement Supérieur et de la Recherche.

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.