Recurrent Rearrangement of the Ewing’s Sarcoma Gene, EWSR1, or Its Homologue, TAF15, with the Transcription Factor CIZ/NMP4 in Acute Leukemia¹

Oncogenic fusions involving the RNA-binding proteins of the TET³ family have been identified in both solid tumors and acute myeloid leukemia and are characterized by a common structure in which the COOH-terminal RNA-binding domain of the TET proteins is replaced by a transcription factor or its DNA-binding domain. The systematic and interchangeable presence of the NH₂-terminal transactivation domain of the different TET proteins suggests a common oncogenic mechanism. The tumor phenotype, however, is defined by the transcription factor. Fusions to Ets factors are found in Ewing’s sarcoma. ERG fusions are found in both Ewing’s sarcoma and acute myeloid leukemia. The C/EBP factor CHOP fusions are responsible for myxoid liposarcoma, and fusions with WT1 are found in desmoplastic round cell tumors, ATF1 in clear cell sarcomas, and TEC/NOR1 in extraskeletal myxoid chondrosarcoma (1). The molecular mechanisms of transformation by TET fusions are not yet resolved. The TET fusions were reported to bind to the same DNA sequences as the wild-type transcription factors and to act as transcriptional activators of reporter constructs carrying those sequences, a property that is dependent on the presence of the TET-transactivating domain (2, 3). EWSR1-FLI1 induces expression of manic fringe (4) and represses expression of TGF-β type II receptor (5), and both events are linked to transformation. Other target genes of EWSR1-FLI1 include MMP3, EAT2, and CK15 (6). However, evidence was presented that also DNA binding domain-independent pathways are involved in EWSR1-FLI1-mediated oncogenesis (7, 8). More recently, it was reported that TET proteins are involved in pre-mRNA splicing, a process that is affected by oncogenic TET fusions (9, 10, 11).

We have collected seven acute leukemia cases with a novel translocation t(12;17)(p13;q11) or its variant t(12;22)(p13;q12) and show that the rearrangements affect the CIZ/NMP4 transcription factor gene on chromosome 12 and the EWSR1 or TAF15 genes on chromosome 17 and 22, respectively. The recurrent TAF15-CIZ and EWSR1-CIZ fusions described here expand the oncogenic properties of TET fusions toward lymphoid malignancies, a phenotype that must be defined by CIZ/NMP4, a transcription factor not yet implicated in hematopoiesis.

Seven patients presented with a diagnosis of acute leukemia: case 1, acute undifferentiated leukemia; case 2, acute myeloid leukemia, FAB-M1; cases 3–6, acute lymphoblastic leukemia with pro-B phenotype (CD19+, CD10−); and case 7, a common B-ALL. Their bone marrow karyotype showed either a t(12;17)(p13;q11) in cases 2–6 or a t(12;22)(p13;q12) in cases 1 and 7 in the majority of the metaphases. The patients were diagnosed in Bordeaux, France (case 1); Nantes, France (cases 2); Italy (cases 3–6); and Leuven, Belgium (case 7).

5′-RACE and RT-PCR experiments were performed according to established protocols. mRNA was reverse transcribed with the CIZ-Ex4R1 (5′-GGTCAGCTGGTCTGACTTGGA) primer. The specific primers used for the 5′-RACE were CIZ-Ex4R2 (5′-TCTGGCAACAGCTGATCCTTC) and CIZ-Ex3R (5′-GAAGGCCAGAAGTACGGGTTAGA). The TAF15-CIZ fusion was confirmed by heminested RT-PCR on patient cDNA using primers CIZ-Ex4R1 and CIZ-Ex4R2, in combination with primer TAF2N-Ex3F (5′-GGCTATGGACAAGCATCACA). The EWSR1-CIZ fusion transcript was detected with CIZ-Ex4R1 and CIZ-Ex4R2 in combination with EWSR1bf (5′-CCCAAACTGGATCCTACAGC).

CIZ was amplified from human bone marrow cDNA with the primers CIZflF1 (5′-GAGATCTAATGCCGGCAGAATGGAAGAATCTCACTTC) and CIZfl-R1 (5′-AGGGCCCAGAGCTGGCCAGGTGCTCCAC); the NH₂-terminal region of TAF15 was amplified with TAF2N-F1 (5′-GAGATCTGCCACCATGTCGGATTCTGGAAGT) and TAF2N-R1 (5′-AGCCGGCATCTGTTCTGGGTCCATAATC); and the NH₂-terminal region of EWSR1 was amplified with primers EWS-F1 (5′-GAGATCTGAGAAAATGGCGTCCACGGATTAC) and EWS-R1b (5′-TGCCGGCCTGCCCGTAGCTGCTGCTCTGT). All PCR products were cloned in pGEM-Teasy, and both strands were sequenced. Expression constructs for CIZ, EWSR1-CIZ, and TAF15-CIZ were constructed in pcDNA3.1–6HysMyc, introducing a COOH-terminal 6His-Myc-tag. The inserts were then moved to pMSCV-puro (a gift from G. Gilliland, Harvard University, Boston, MA).

HEK293T cells were transiently transfected with the pMSCV plasmids using the Fugene 6 Transfection Reagent (Roche) and lysed after 24 h as described (12). DNA probes were prepared by annealing the oligonucleotides 5′-CGCGTCAACCTTTTTCAAAAAGACCAG and 5′-CTAGCTGGTCTTTTTGAAAAAGGTTGA with the −320 to −305 region of the human MMP1 promoter and filling in the 5′-overhangs with Klenow polymerase and [α-³²P]dCTP. Mutant probe was prepared by annealing the oligonucleotides 5′-CCTGTGTCAGAGAGA and 5′-TCTCTCTGACACAGG.

The −296HMAT-Luc reporter construct with the human MMP7 promoter in pGL2Basic was a gift from L. Matrisian (Vanderbilt University, Nashville, TN; Ref. 13). HEK293T cells at 35% confluence were cotransfected with the reporter construct (150 ng), together with each of the expression vectors (or empty pMSCV-puro) and 15 ng of a β-galactosidase expression plasmid, using FuGene-6 Reagent (Roche). After 36 h, cells were lysed in passive lysis buffer (Promega) and assayed for luciferase activity using a commercial assay (Promega). All values were normalized for β-galactosidase activity. Experiments were performed three times, each point in triplicate.

Viral supernatants were produced in HEK293T cells cotransfected with the pMSCV-puro constructs and the pIK6.1MCV packaging vector (a gift from G. Gilliland, Harvard University). NIH3T3 cells were seeded at 2 × 10⁵ cells/well of a 6-well plate 24 h before adding the viral supernatants containing 8 μg/ml Polybrene for 4 h. Twenty-four h later, selection with puromycin (2 μg/ml) was started. For the focus formation assay, the transduced NIH3T3 cells were seeded at a density of 10⁶ cells per 9-cm culture dish in DMEM/10% FCS and cultured for 3 weeks. Medium was replaced every 3.5 days. Soft agar assays were performed as described (7).

RNA was isolated form NIH3T3 or Ba/F3 cells expressing the different fusion genes and treated with DNaseI (Invitrogen). RT-PCR was performed with oligonucleotides 5′-GAGGAGCTAGCAGGTTATCCTAA and 5′-AGCTACACAGTGCTTCTGAACATC for Mmp3, 5′-AAGGAGAGATCATGGAGACAGCTT and 5′-GATGTCTCGCAACTTCATGTTACC for Mmp7, 5′-CCCTCAGCTACGGTGTCT and 5′-TCGTAGCCTTTCCTGTCA for manic fringe, and 5′-CTGACCAAGCGAGAGTGTGA and 5′-TCCCTCTTTGGCATAGGTTG for Eat2. Thirty-five cycles of denaturation for 30 s at 95°C, annealing for 30 s at 56°C, and extension for 40 s at 72°C were performed. The EWSR1/FLI1 expression plasmid was obtained from C. Denny (UCLA, Los Angeles, CA).

Metaphases of an acute myeloid leukemia with t(12;17)(p13;q11) and an acute undifferentiated leukemia with t(12;22)(p13;q12) as the sole anomaly were analyzed by fluorescence in situ hybridization using bacterial artificial chromosome clones mapped to chromosome region 12p13. Clone 433J6 (accession number AC079387) showed split signals in both cases. The 12p breakpoint was narrowed by Southern blot experiments, which, combined with the sequence from 433J6, indicated that a gene highly similar to rat Ciz (p130CAS interacting zinc finger, rCIZ; accession number AB019281) was disrupted by both translocations. rCIZ was identified as a p130CAS interacting protein (14) and is identical to the NMP4 (15). Next, we obtained two cDNAs of 1.55 and 1.74 kb, respectively, for human CIZ by RT-PCR on bone marrow cDNA. The longer cDNA has an open reading frame coding for 578 amino acids showing 92% identity to rCIZ and is identical to the human CIZ cDNA sequence reported recently (accession number XM_033227; Fig. 1 A). The shorter transcript is a splice variant coding for a 517 amino acid protein, similar to splice variants described for rCIZ/NMP4 (14, 15). The gene has 16 exons spanning ∼60 kb of genomic DNA.

To identify the fusion partner of CIZ, 5′-RACE experiments were performed on cDNA from the t(12;17) case. A fusion fragment linking exon 6 of the TAF15 gene (also known as RBP56 or TAF2N) to exon 3 of CIZ was detected (Fig. 1, B and C). This TAF15-CIZ rearrangement was confirmed by RT-PCR and results in an in-frame fusion of the 5′ part of TAF15, encoding its transactivating domain, to the entire CIZ sequence. TAF15 is a member of the TET family of proteins, which also includes EWSR1 and TLS/FUS. We thus investigated whether the t(12;22) involved EWSR1. Using RT-PCR with EWSR1 and CIZ primers, an in-frame fusion linking exon 7 of EWSR1 to exon 2 of CIZ was detected. As the start codon of CIZ is located in exon 3, this fusion protein contains some novel sequences derived from the 5′ UTR of CIZ, as well as the complete CIZ amino acid sequence (Fig. 1, B and C). Attempts to amplify possible reciprocal transcripts were negative.

Four additional leukemias with a t(12;17) were collected; interestingly, these were all ALLs. A TAF15-CIZ fusion cDNA was detected by RT-PCR cDNA in the 3 cases for which cDNA was available. The structure of the fusion was identical to the one described above for two cases, whereas in the third case, a fusion of exon 9 of TAF15 to exon 3 of CIZ was found (Fig. 1, B and C). Fluorescence in situ hybridization analysis detected a break within the CIZ gene in the case for which no RNA was available, strongly suggesting that this also harbors a TAF15-CIZ rearrangement. One additional case, also an ALL, with a t(12;22) was found. RT-PCR detected a variant EWSR1-CIZ fusion (EWSR1 exon 7-CIZ exon 3; Fig. 1, B and C).

These results show that the TET-CIZ fusions are a recurrent rearrangement, mainly in ALL, but also occurring in acute myeloid leukemia. TET fusions were not described previously in acute lymphoblastic malignancies, and this novel phenotype must be defined by the COOH-terminal partner, CIZ. Bone metabolism is the only physiological process in which CIZ was implicated to date, despite its rather ubiquitous expression (14, 15). Our data indicate that CIZ is also implicated in normal and aberrant lymphoid and myeloid development.

The oncogenic properties of the TET-CIZ fusions were analyzed by overexpression in NIH3T3 and Ba/F3 cell lines. NIH3T3 cells stably expressing TAF15-CIZ or EWSR1-CIZ displayed a clearly transformed phenotype, compared with cells transduced with empty pMSCV-puro vector or cells overexpressing CIZ (Fig. 2). The cells are smaller, grow to a higher density, and form multiple foci. Two distinct populations were observed: cells with a small round cell morphology and spindle-like cells with a “herring-bone” appearance, reminiscent of fibrosarcoma. Interestingly, a similar round cell phenotype was described for EWSR1-FLI1 and EWSR1-ETV1 (16). Controls with constructs expressing exclusively the EWSR1 or TAF15 part of the fusion, or wild-type CIZ, resulted in a normal phenotype (not shown), demonstrating that the transformation properties of the fusions are dependent on the presence of both the NH₂-terminal domain of TAF15 or EWSR1 and CIZ, respectively. Expression of TAF15-CIZ or EWSR1-CIZ, however, could not induce substrate-independent growth of NIH3T3 cells in soft agar (whereas a control with BCR-ABL expressing RAT-1 cells was positive; result not shown). It was shown previously that EWSR1-FLI1 supports anchorage-independent growth of NIH3T3 cells, but that EWSR1-ETV1 does not, showing that soft agar growth is a transformed phenotype not shared by all TET fusions (16). This shows clearly that the transcription factor determines the strength of transformation and suggests that CIZ defines the weak transforming potential of the TET-CIZ fusions.

Expression of the fusion constructs in the interleukin 3-dependent murine pre-B lymphoid cell line Ba/F3 failed to induce growth factor independence or increased resistance to apoptosis (not shown).

CIZ has the properties of a transcription factor: it localizes in focal adhesions and in the nucleus of adherent cell lines, it binds specifically to the (G/C)AAAAA(A) sequence, and overexpression of CIZ was shown to activate MMP1, MMP3, and MMP7 promoters that contain the CIZ consensus sequence (14). To investigate the transactivating properties of TET-CIZ fusions, EMSAs were performed with extracts of HEK293T cells transiently expressing myc-tagged TAF15-CIZ or EWSR1-CIZ. These showed specific binding of the chimeric proteins to DNA with the CIZ consensus binding sequence; a retarded labeled band is observed that can be supershifted with antibodies against the myc-tag and disappears upon addition of unlabeled probe (Fig. 3). Next, reporter assays were performed. Overexpressed CIZ in HEK293T cells had a weak but reproducible transactivating activity on a luciferase reporter construct carrying the −301 to +35 region of the human MMP7 promoter as described previously (14). Surprisingly, TAF15-CIZ and EWSR1-CIZ and CIZ did not show enhanced (or decreased) transactivation of this promoter at similar expression levels (Fig. 4). We were not able to increase the expression levels of the EWSR1 and TAF15 fusions, possibly because these interfere with proliferation of the cells, as described for EWSR1-FLI1 (17).

It is puzzling that the TET-CIZ fusions do not transactivate reporter constructs carrying CIZ binding sites, although they do bind to the CIZ DNA recognition sequence in EMSA experiments. It is possible that HEK293T cells lack some CIZ- or EWSR1/TAF15-interacting factors, preventing the full transactivating activity of the TET-CIZ proteins. However, it should be noted that HEK293T cells do express endogenous CIZ, and that increasing the concentration of CIZ is able to enhance the activation of the MMP7 promoter.

The transactivation properties of the TET-CIZ fusions were thus further explored, and the transcriptions of the CIZ target mmp7 and of the EWSR1-FLI1 targets mmp3, eat2 (6), and manic fringe (4) were analyzed in NIH3T3 and BA/F3 cells by RT-PCR (Fig. 4). No transcription of any target was seen in NIH3T3 cells expressing either TET-CIZ fusion, whereas a strong induction of mmp3, eat2, and manic fringe transcription was observed in the EWSR1-FLI1-expressing controls. Similarly, no induction of mmp3 or mmp7 was visible in the Ba/F3 cells expressing the TET-CIZ fusions. Interestingly, eat2 and manic fringe were found to be constitutively expressed in this cell line. The semiquantitative character of the RT-PCR did not allow us to determine whether the transcription of these genes was increased.

Taken together, these results suggest that the CIZ fusions do not act as transcriptional activators (or repressors) for the MMP3 or MMP7 promoters, although they do bind the CIZ consensus sequence, or that MMPs are not the true targets of CIZ. On the other hand, it has been shown that some of the transforming activities of EWSR1-FLI1 fusions are not dependent on DNA binding and transactivation (7, 8). In this regard, the observation that TET proteins do affect pre-mRNA splicing (11, 18), a process that is affected by EWSR1-FLI1 (11, 19), TLS-ERG (9), and EWSR1-NOR1 (10), is of particular interest. Alternative splicing of particular genes can lead to modified functional properties of the gene products, e.g., dominant-negative isoforms of ikaros and a decrease of ikaros activity is associated with blast crisis in chronic myeloid leukemia (20). If transcriptional activation is not the major mechanism of action of TET-CIZ fusions, these could provide an interesting model to investigate alternative oncogenic pathways for TET fusions, such as the effect on pre-mRNA splicing.

We thank Jan Cools for thoughtful comments and Drs. G. Specchia, E. Miraglia, J. Reiffers, P. Cony-Makhoul, and A. Uyttebroeck for patient samples.