PLAG1 Fusion Oncogenes in Lipoblastoma¹

Lipoblastomas are benign adipose tumors that are circumscribed(lipoblastoma) or diffuse and infiltrative (lipoblastomatosis). They are usually diagnosed in children under the age of 5 years and are thought to arise from lipoblastic cells resembling those in the embryonic and fetal stages of development (1, 2, 3). However,unlike the benign lipomas of adulthood, they can grow rapidly and may recur after resection. Most lipoblastomas arise in the soft tissues of the extremities and are more prevalent in males (1, 2, 3). They are distinctive histologically and are classically composed of lobulated tissue containing a variable number of lipoblasts and stellate mesenchymal cells as well as immature and mature adipocytes. Other typical findings include a plexiform capillary network and a prominent myxoid matrix consisting primarily of mucopolysaccharides(1). Although the pathogenesis of lipoblastoma is uncertain, it is believed to be more closely related to developing white fat than brown fat (4, 5, 6, 7). Given the resemblance of lipoblastoma cells to those in embryological and fetal fat, it has been suggested that the biological pathways responsible for lipoblastoma proliferation might be related to those involved in adipose tissue development (8). Hence, characterization of oncogenic mechanisms in lipoblastoma might shed light on biological pathways that regulate proliferation and maturation of developing adipose tissue.

All human lipoblastomas reported to date contain clonal chromosomal rearrangements involving the 8q11–13 region (9, 10, 11, 12, 13, 14, 15). In contrast, 8q11–13 rearrangements are uncommon in other types of adipose tumors, including lipoma and liposarcoma. The lipoblastoma 8q11–13 rearrangements are generally balanced, involving no net loss or gain of 8q material as judged at the cytogenetic level of resolution. The balanced nature of these rearrangements is most consistent with an oncogenic activating mechanism in which one or more genes in the 8q11–13 region are rearranged and/or transcriptionally up-regulated. Notably, there appear to be varied mechanisms by which the putative lipoblastoma 8q oncogene is activated. Reported lipoblastoma 8q rearrangements include fusions with the telomeric aspect of the same chromosome arm (10) and translocations with a number of different partner chromosomes (9, 11, 12, 13, 14, 15). Therefore, the cytogenetic evidence implicates an 8q11–13 oncogene in the neoplastic transformation of immature fat cells. However, the nature of that gene and of the various 8q11–13 partner genes is unknown.

To characterize molecular mechanisms of adipose cell transformation, we mapped and cloned the oncogenes associated with lipoblastoma chromosome 8q rearrangements. Structurally similar PLAG1 fusion genes were identified in each of four lipoblastomas. These studies reveal a major oncogenic mechanism in lipoblastoma and suggest biological pathways that might be important in adipose tissue development.

YAC clones were obtained from Research Genetics (Huntsville, AL) or CEPH (Paris, France). YAC DNA was isolated as described previously(16), except that YAC cultures were grown in bulk and harvested directly. BAC clones were isolated from the Research Genetics CITB Human BAC DNA Pools Release IV using primers PLAG 1650-E5-F(CCCTAGATGATGGTGCAGGAGACCTC) and PLAG 1873-E5-R(CTGAAGATCCTGTGTTTGTGTGG). BAC DNA was isolated using an established protocol (4), with minor modifications. Bacterial cell pellets were frozen at −80°C, thawed, and then digested with 25 mm Tris-HCl (pH 8.0), 50 mm glucose, 10 mm EDTA, 2.5–5.0 mg/ml lysozyme, and 100–200 μg/ml RNase. DNA was precipitated by the addition of 0.1 volume of 3 m NaOAc and 2 volumes of 95% ethanol. The presence of PLAG1 in BAC227K20 was confirmed by PCR amplification using primers PLAG 17-E1-F (CAATGGCTGCTGGAAAGAGG) and PLAG 133-E1-R (CCCGTCCGCCG CCTCTACACC).

Metaphase harvest and slide preparation were performed as described previously (17), with minor modifications. Metaphase cell preparations were applied to glass slides, which were then stored at room temperature for up to 2 weeks and pretreated by incubation in 2×SSC for 1 h at 37°C before hybridization. YAC and BAC clones were biotin or digoxigenin labeled by random octamer priming using the BioPrime DNA Labeling System (Life Technologies, Inc., Rockville, MD). The labeled products were purified by S-200HR spin column chromatography (Pharmacia, Uppsala, Sweden), coprecipitated with 5×Cot-1 DNA and 1 μg of herring sperm DNA, and then resuspended in hybridization buffer (50% formamide, 10% dextran sulfate, and 2×SSC). Dual-color FISH with BAC, YAC, and centromeric probes was performed as described previously (16), except that washing steps were performed using PBS-T (PBS with 0.1% Tween 20). Probe detection was performed in a stepwise manner as follows: slides were incubated with FITC antidigoxigenin and Texas Red-streptavidin,then incubated with anti-FITC rabbit IgG, then incubated with biotinylated antistreptavidin, and, finally, incubated with FITC antirabbit and Texas Red-strepavidin (Zymed Laboratories, San Francisco, CA). All detection incubations were for 30 min at 37°C,and slides were washed three times in PBS-T after each incubation. Images were obtained using a cooled charge-coupled device camera(Photometrics).

Total RNA was isolated using Trizol (Life Technologies, Inc.) according to the supplier’s protocol after homogenization of tissues by extensive vortexing. 5′ RACE was performed using the Marathon cDNA amplification kit (Clontech Laboratories, Inc., Palo Alto, CA)according to the manufacturer’s instructions, with minor modifications as described below. First-strand cDNAs were synthesized in a 3-h reaction from 5 μg of total RNA. Amplification of 5′ cDNA ends was performed using adapter primer AP1 and PLAG1 exon 5 primer MV5 (18). Nested PCR amplification was performed using adapter primer AP2 and PLAG1 exon 5 primer MV6(18). All amplifications were performed using Advantage 2 Polymerase Mix (Clontech Laboratories, Inc.). Gel-purified nested PCR products were cycle sequenced by incorporation of ABI PRISM Big Dye Terminators (Perkin-Elmer) and analyzed on an ABI 377 automated sequencer.

RT-PCR was performed using the Gene Amp RNA PCR Kit (Perkin-Elmer) with first-strand synthesis using either an oligodeoxythymidylic acid primer or the PLAG1 primer MV5 (18). HAS2-PLAG1 fusion cDNA was then amplified by nested PCR using Advantaq Plus (Clontech Laboratories, Inc.). First-round primers were HAS 420-E1-F (GTCGTCTCAAATTCATCTGATCTC) and PLAG 433-E4-R(TCTTGTTGG ACACTTGGGAAC), and second-round primers were HAS 502-E1-F(CTGAGGACGACTTTATGACCAG) and PLAG 415-E4-R(CTTTAGGTGGCTTCTCAAGTTTC). The COL1A2-PLAG1 fusion was amplified using primers COL1A2 E1+112F (AAGGAGTCTGCATGTCTAAGTGCTA) and PLAG 415-E4-R.

Total RNA was isolated from cultured cells using Trizol (Life Technologies, Inc.), separated by electrophoresis through a formaldehyde-containing gel, and then transferred to a Hybond-N membrane (Amersham Pharmacia Biotech). The blot was hybridized with a PLAG1 cDNA probe (2074 bp) obtained by nested PCR using primers PLAG 947-E5-F (TGCAAACTTTTGAAAGCACG) and PLAG 3928-E5-R(ATGAAGTGCGGTATGTGTGC), followed by PLAG 1265-E5-F(ATAAAAGACGAGCTCCTTCCG) and PLAG 3338-E5-R (CAGAGATGCATGAAA GTGGG). The blot was then stripped and rehybridized with a HAS2 cDNA probe (785 bp)obtained using primers HAS 49-E1-F (CCCATTGAACCAGAGACTTGAAA) and HAS 833-E2-R (GTTCAACTTTATGGGGGTTTCTA). All cDNA probes were labeled using a Prime-It II Random Primer Labeling Kit (Stratagene, La Jolla, CA) and then purified by S-200HR spin column chromatography (Pharmacia).

Metaphase cell FISH mapping revealed that CEPH YAC 166f4 spanned the 8q12 breakpoints in lipoblastoma cases 1–4. YAC 166f4 is known to contain the PENK (preproenkephalin), MOS, and PLAG1 genes. Notably, PLAG1 is an oncogene targeted by chromosomal translocations in salivary gland tumors(18). To evaluate PLAG1 as a potential target of the lipoblastoma 8q12 breakpoint, we next identified a BAC clone containing sequences from the entire PLAG1 coding sequence. This clone, 227k20, spanned the lipoblastoma 8q12 breakpoints (Fig. 1,A). We then performed FISH mapping to identify a CEPH YAC clone, 947h7, spanning the 8q24 breakpoint in lipoblastoma case 1 (Fig. 1 B). This 1400-kb YAC contains the HAS2 gene encoding hyaluronan synthase type 2 and also contains the Langer-Gideon critical region, including the loci for TRPS1(tricho-rhino-phalangeal syndrome 1) and EXT1 (exostoses 1).

Because FISH mapping implicated PLAG1 as a strong candidate gene, we performed RACE to identify potential PLAG1 fusion partners. In case 1, PCR amplification of the 5′ end of the PLAG1 transcripts gave 950- and 1050-bp products. Sequence analysis of these two products revealed an identical ectopic sequence fused to PLAG1 exon 3 (GenBank accession number AF221548) or exon 2 (GenBank accession number AF221549), respectively. This finding indicates a genomic fusion breakpoint in PLAG1 intron 1 with alternative splicing of exon 2 and is consistent with a known splice variant of PLAG1 (18). BLAST analysis identified the ectopic sequence as exon 1 of the hyaluronic acid synthase gene, HAS2 (Fig. 2 A). HAS2 maps to chromosome 8q24.1(19) and to YAC clone 947h7 that spans the 8q24.1 breakpoint in lipoblastoma case 1. In addition, Southern blotting analyses revealed genomic PLAG1 and HAS2rearrangements in lipoblastoma case 1 (data not shown). Hence, both genomic and mRNA analyses in lipoblastoma case 1 support HAS2-PLAG1 fusion as the oncogenic consequence of the cytogenetic 8q12:8q24.1 fusion event.

RACE analysis of lipoblastoma case 4 revealed a second PLAG1fusion partner. This lipoblastoma contained an apparently balanced translocation, t(7;8). PCR amplification of the PLAG1 5′ end gave a 617-bp product, and sequence analysis revealed a fusion transcript in which exon 1 of the COL1A2 gene was joined to exon 3 of PLAG1 (GenBank accession number AF221550; Fig. 2 A). COL1A2 maps to chromosome band 7q22.1, and its fusion with PLAG1 is consistent with the cytogenetic mechanism in lipoblastoma case 4.

RT-PCR analyses in the group of four lipoblastomas demonstrated HAS2-PLAG1 fusions in cases 1–3, whereas COL1A2-PLAG1 fusion was found only in case 4 (Fig. 2,B). Each of the cases with HAS2-PLAG1 fusions had intrachromosomal 8q rearrangements, which were 8q12:8q24.1 fusions in cases 1 and 2 and a ring chromosome 8 in case 3 (Table 1). Two HAS2-PLAG1 fusion products (293 and 188 bp) were identified in cases 1–3, and sequence analysis revealed that these products were alternative splicing variants that either included or lacked PLAG1 exon 2, respectively. Similarly, COL1A2-PLAG1 fusion products (354 and 249 bp) were alternative splicing variants that differed only in the inclusion of PLAG1 exon 2. Therefore, the genomic translocation breakpoints are within PLAG1 intron 1 in each of the four lipoblastomas (Fig. 2 B).

PLAG1 expression was evaluated by Northern blot analysis in a HAS2-PLAG1 lipoblastoma cell line (established from case 1) and in cell lines established from other benign and malignant mesenchymal tumors (Fig. 3,A). A 7.5-kb transcript similar in size to the known wild-type PLAG1 transcript was expressed strongly in the lipoblastoma. Rehybridization with a 785-bp probe containing HAS2 exons 1 and 2 confirmed the fusion nature of the lipoblastoma 7.5-kb PLAG1 transcript. HAS2transcripts of 3.2 and 4.4 kb were demonstrated in all tumors, whereas an additional 7.5-kb transcript was detected exclusively in the lipoblastoma (Fig. 3 B). The 7.5-kb HAS2-PLAG1fusion transcript was not detected when the blot was rehybridized with a 1389-bp HAS2 cDNA probe lacking exon 1 (data not shown).

In previous studies, we characterized two lipoblastomas (Table 1,cases 1 and 2,) that had 8q intrachromosomal rearrangements joining band 8q12 with 8q24.1 (10). The intervening material, from chromosome band 8q12 to 8q24.1, was inserted into another chromosome in each case. We hypothesized that the 8q12::q24.1 fusions were associated with juxtaposition of genes from those two regions. In this study we demonstrate that PLAG1, a developmentally regulated zinc finger gene, is the oncogene target in lipoblastoma chromosome band 8q12 rearrangements. PLAG1, which encodes a transcription factor comprised of seven NH₂-terminal C₂H₂ zinc fingers and a COOH-terminal serine-rich region, was discovered recently by virtue of its oncogenic role in pleomorphic adenomas of the salivary gland(20). We now report novel PLAG1 fusion mechanisms involving the HAS2 gene at chromosome band 8q24.1 and the COL1A2 gene at chromosome band 7q22 in lipoblastoma. Additionally, we show that PLAG1 is expressed more strongly in HAS2-PLAG1 lipoblastoma cells than in several other mesenchymal tumors.

The mechanism of PLAG1 oncogenic activation in pleomorphic salivary adenomas differs from that in most fusion oncogenes. Pleomorphic adenoma translocation breakpoints generally involve the 5′untranslated regions of both PLAG1 and a fusion partner gene. Consequently, the entire PLAG1 coding sequence, which begins in exon 4, is placed under the transcriptional control of an active and ectopic promoter region. These “promoter-swapping”events lead to high-level PLAG1 expression in salivary gland cells, where PLAG1 transcripts are normally undetectable. The most frequent PLAG1 fusion in pleomorphic adenomas involves promoter swapping with the constitutively expressedβ-catenin gene, CTNNB1 (18). Alternate tumorigenic mechanisms involve PLAG1 promoter swapping with the constitutively expressed genes for leukocyte inhibitory factor receptor (LIFR; Ref. 21) and transcription elongation factor A 1 (TCEA1; Ref. 22).

Here we implicate PLAG1 promoter swapping as a critical event in lipoblastoma tumorigenesis. Notably, the general mechanism of PLAG1 activation in lipoblastoma is similar to that in pleomorphic adenomas. Most pleomorphic adenomas, like the lipoblastomas reported herein, have chromosomal breakpoints involving PLAG1 intron 1 and resulting in loss of PLAG1exon 1. A hypothetical explanation for the recurrent PLAG1intron 1 breakpoints is that upstream sequences might contain transcriptional repressor binding sites. If so, the intron 1 breakpoints would enable PLAG1 transcriptional up-regulation by introducing active, ectopic promoter regions and by eliminating negative control elements.

Although PLAG1 mechanisms in pleomorphic adenomas and lipoblastomas are likely related, the specific PLAG1 fusion partners differ in these two tumors. We identified HAS2-PLAG1 or COL1A2-PLAG1 fusion genes in each of four lipoblastomas, whereas those fusions have never been reported in pleomorphic adenoma. In addition, published reports do not reveal lipoblastomas having the specific translocations, including t(3;8) and t(5;8), that frequently activate PLAG1 in pleomorphic adenomas (19, 21, 22).

The genomic HAS2-PLAG1 fusion breakpoints are in HAS2 intron 1, whereas the HAS2 coding sequence begins with the first codon of exon 2. Therefore, the HAS2-PLAG1 fusion genes contain the entire HAS25′ untranslated region. There are no known transcription factor binding sites in the HAS2 coding sequences, and it is therefore likely that HAS2-PLAG1 is under the transcriptional control of an intact HAS2 promoter. Histological correlations suggest that the HAS2 promoter might be particularly active in lipoblastoma cells. HAS2 belongs to a multiprotein family that controls hyaluronic acid production at the level of either synthesis or transport from the cell (23, 24). Hyaluronic acid is a glycosaminoglycan that influences various biological activities,including cellular maturation, migration, and metabolism. Hyaluronic acid synthesis and metabolism are tightly regulated (25),and lipoblastomas have a variably myxoid stroma containing extracellular hyaluronic acid (1, 26). These observations are consistent with the notion that HAS2 is expressed at relatively high levels in lipoblastoma cells. If so, lipoblastoma cell HAS2 promoter swapping might be a particularly effective mechanism for accomplishing PLAG1 up-regulation.

In addition to the HAS2-PLAG1 fusion, we have characterized a second lipoblastoma fusion involving the PLAG1 and COL1A2 genes. The COL1A2-PLAG1 fusion gene retains a short portion of the COL1A2 coding sequence in exon 1 that is joined with either PLAG1 exon 2 or exon 3 in the fusion transcripts. Consequently, COL1A2-PLAG1 encodes a full-length PLAG1 protein and a short, COOH-terminal-truncated, COL1A2 protein. The truncated COL1A2 protein is comprised of the first 23 amino acids of COL1A2, followed by 25 or 22 additional amino acids encoded by PLAG1 exons 2 or 3, respectively. However, it is unknown whether this predicted COL1A2-PLAG1 fusion protein is functional or stably expressed. COL1A2, a member of the collagen gene family, encodes a protein that coils at a 1:2 ratio with collagen 1 α 1 to form the collagen fibers that make up bones,tendons, and connective tissues (27). Notably,lipoblastomas are often arranged in nodules separated by collagenous septae (26). It is possible that the lipoblastoma cells are directly responsible for production of the collagenous architectural features. Hence, COL1A2 might be an especially strong promoter in lipoblastoma cells.

Our findings in lipoblastoma provide the first example of PLAG1 rearrangement in mesenchymal neoplasia. PLAG1 expression has been demonstrated previously in certain mesenchymal tumors, particularly those of smooth muscle cell origin(22), but PLAG1 rearrangements have been found only in pleomorphic adenomas, which are believed to be epithelial neoplasms. Here we demonstrate stronger PLAG1 expression in HAS2-PLAG1 lipoblastoma cells than in several other benign or malignant mesenchymal tumors. Lipoblastoma PLAG1expression was 4- and 6-fold greater than that in lipoma and malignant peripheral nerve sheath tumor, respectively, whereas PLAG1transcripts were undetectable in an osteosarcoma. These findings provide further evidence that tumor chromosomal rearrangements affect PLAG1 transcriptional up-regulation. Most lipoblastomas have cytogenetic rearrangement of the PLAG1 (chromosome band 8q12) region, and it is likely that PLAG1 activation is a pivotal oncogenic event in lipoblastoma.

Notably, PLAG1 transforming activity in lipoblastoma may be related to its recently described function as a transcriptional activator of the IGF2 gene. IGF2 is up-regulated by PLAG1 interactions at an IGF2 promoter 3 consensus binding site, and IGF2 is expressed abundantly in salivary gland adenomas containing PLAG1 oncogenes (28). IGF2 stimulates proliferation of adipocyte progenitor cells (29), and it is therefore likely that IGF2 transcriptional up-regulation contributes to neoplastic proliferation in PLAG1-mutant lipoblastoma cells.