Novel SH3 protein encoded by the AF3p21 gene is fused to the mixed lineage leukemia protein in a therapy-related leukemia with t(3;11) (p21;q23)

A 14-year-old female was diagnosed as having T-cell ALL in 1988. The karyotype of the leukemic blasts was 46/XX. She was treated according to a protocol for high-risk ALL of Osaka Children's Leukemia Study Group, followed by an autologous peripheral blood stem cell transplantation. The cumulated doses of antileukemia agents used were: 300 mg/m² daunorubicin, 2600 mg/m2etoposide, 7500 mg/m2 cyclophosphamide, and 460 mg/m² ranimustine. The patient had been well until she developed AML (FAB M5b) in November 1997. The karyotype of the t-AML blasts was t(3;11) (p21;q23). Complete remission was achieved with a Koseisho protocol for acute nonlymphoblastic leukemia (ANLL-91), and in September 1998, she underwent bone marrow transplantation with cells donated from a human lymphocyte antigen–identical unrelated donor. Since that time, she has been in remission with chronic graft-versus-host disease.

A genomic library from the patient's leukemic cell DNA was constructed by the complete digestion of genomic DNA with BamH I and ligation into a phage vector (λZAP Express; Stratagene, La Jolla, CA). Rearranged BamH-I fragments were isolated by screening with theMLL complementary DNA (cDNA) probe.

Total RNA extracted from the leukemic cells was subjected to reverse transcriptase–polymerase chain reaction (RT-PCR) usingMLL exon 5 (forward) and AF3p21 (reverse) primers for amplification of MLL-AF3p21 fusion transcripts. The sequences of the primers were: forward, 5′-AAGCAGCCTCCACCACCAGA, and reverse, 5′-GCTGGAGCCTGGTTCAGATG. The PCR products were subcloned into a plasmid vector (pGEM-T Easy vector; Promega, Madison, WI) and sequenced.

To analyze the involvement of the MLL gene in this translocation, we performed Southern blot analysis of DNA prepared from the leukemic cells of the patient. Hybridization of BamH-I digested DNA with the MLL cDNA probe revealed the presence of approximately a 4-kilobase (kb) rearranged band in addition to normal-size fragments (data not shown). To clone the rearranged DNA, we constructed a genomic DNA library prepared from BamH-I digested leukemic cell DNA. Screening of the library identified 2 clones composed of a rearranged MLLsegment among 13 positive clones. Sequence comparison between these clones and the 8.3-kb MLL breakpoint cluster region (bcr) revealed that MLL sequences downstream of exon 6 were substituted with the non-MLL DNA segment.

A BLAST database search for the novel sequence identified a highly homologous expressed sequence tag (EST) clone (GenBank accession no. AA227 226). Since only a portion of the 5′ terminus sequence of AA227 226 was deposited in the database, we determined the entire sequence of the EST clone. This clone contained a 3003–base pair (bp) insert, and a reading frame analysis suggested that this clone possibly contains an extraneous sequence in its 5′ region. Therefore, we performed 5′ RACE (rapid amplification of cDNA ends) to isolate the major spliced products and to determine the transcription start site of the novel gene using a cDNA of K562 leukemia cell line as a template. This analysis revealed that AA227 226 indeed contains a 29-bp intronic sequence at the exon-intron junction of exon 2 and lacks the 49-bp sequence at its 5′ terminus. The corrected sequence has 3023 bp with an open reading frame of 2166 bp encoding a polypeptide of 722 amino acids with a predicted molecular mass of 80 kd. In keeping with the nomenclature for MLL fusion partner genes, we designated this gene AF3p21 (ALL-1 fused gene from chromosome 3p21). The complete sequence of AF3p21 has been deposited with the National Center for Biotechnology Information (NCBI) Data Library (accession no. AF 178432).

Sequence comparisons with the public database revealed that the N-terminus of the predicted AF3p21 protein (amino acid [aa] 1-58) contains an SH3 domain. The SH3 domain of AF3p21 is mostly similar to that of Fyn proto-oncogene tyrosine kinase. A database search for other functional domains revealed the presence of a bipartite NLS (aa 171-192) and a proline-rich domain (aa 170-249) downstream of the SH3 domain.

To determine the nucleotide sequence around the breakpoint of the chromosome 3p21 region, bacterial artificial chromosome (BAC) DNA containing the AF3p21 gene (clone 425p07; Genome Systems, St Louis, MO) was directly sequenced. A sequence comparison of the genomic segments of MLL, AF3p21, and MLL-AF3p21 at the breakpoint is shown in Figure 1A. There is an identical stretch of 3 nucleotides (CCA) in the MLL andAF3p21 genes at the rearranged site. Therefore, the breakpoint of the MLL gene is located between nucleotide positions 1587 and 1589 (numbering according to the 11q23-breakpoint sequence in GenBank accession no. HSU04 737). This is within an Alu-Sbrepetitive sequence. The breakpoint on chromosome 3 is located at 61 bp upstream of the putative transcription start site of the AF3p21gene.

To determine whether the translocation results in an in-frame fusion of MLL to AF3p21, RT-PCR was performed to amplify the cDNA junction of leukemic cells. Sequence analysis of the chimeric transcripts revealed an in-frame fusion between MLL exon 6 andAF3p21 exon 2 (Figure 1B). The resulting MLL-AF3p21 fusion protein consists of AT-hooks, methyltransferase, and transcription repressor domains of MLL in addition to the AF3p21 proline-rich domain and bipartite NLS but not the AF3p21 SH3 domain (Figure 1C).

Topoisomerase II inhibitor–associated t-AML is characterized by chromosomal translocations, most of which disrupt introns within an 8.3-kb bcr between exons 5 and 11 of the MLL gene. Several translocation breakpoints in MLL and in the partner DNAs are in intronic Alu repeats, suggesting that homologous repetitive sequences may be of importance in joining MLL with its translocation partners. The breakpoint in the MLL gene of our case occurs in an Alu-Sb sequence in intron 6. Although the breakpoint region of the AF3p21 gene does not contain anAlu repetitive sequence, both MLL and AF3p21contained a homologous CCA sequence at the breakpoint (Figure 1A). These results suggest that the base pairing of homologous DNA ends ofMLL and AF3p21 may be a first step in the translocation event in this case.

MLL fusion partner genes encode protein products of different types, and most of them are putative transcription factors.4 There are 2 genes, includingEEN on band 19p13⁹ and ABI-1on 10p11.2,10 that encode SH3 domain-containing protein. EEN is a human homologue of a member of a recently described murine SH3 domain-containing protein family.11 The SH3 domain in EEN is most similar to that found in the GRB2 family proteins, suggesting that it acts as an adapter protein in signal transduction. ABI-1 is a human homologue of mouse Abi-1 (Abl-interactor-1) that suppressesv-Abl transforming activity by binding to the Abl protein.12 The SH3 domain of Abi-1 is responsible for its binding to c-Abl. Both EEN and ABI-1 have an SH3 domain in their C-terminus, and fusion proteins with MLL retain their SH3 domains. However, the SH3 domain in AF3p21 is located at the N-terminus, and the resulting MLL-AF3p21 fusion protein does not retain the SH3 domain (Figure 1C).

Assuming that the MLL partner protein is required for leukemogenesis, the region downstream of the SH3 domain of AF3p21 may have functionally important domain(s). AF3p21 has a bipartite NLS downstream of the SH3 domain and a proline-rich domain. Since the N-terminus of MLL protein retains an NLS,13 the bipartite NLS of AF3p21 may not be required for nuclear localization of the MLL-AF3p21 fusion protein. Proline-rich domain is found in a number of SH3-binding proteins, and this region is responsible for physical interaction between SH3 proteins and their ligands.14 AF3p21 contains 1 class I SH3-binding motif (RXXPXXP) at aa 176-182 and 2 class II motifs (XPPXPXR) at aa 170-176 and aa 242-249. These domain structures suggest that AF3p21 might interact with other proteins in the nucleus. However, the precise intracellular localization and identification of interacting proteins (if any) of AF3p21 await further study.

Chromosomal abnormalities involving 3p21 without involvement of the 11q23 region have been reported in t-AML and t-MDS.15 Moreover, the short arm of chromosome 3 is thought to contain multiple tumor suppressor genes because 1 copy of this chromosomal arm frequently is missing in carcinomas that have arisen in a variety of tissues, and region 3p21.3 is frequently deleted in lung cancers.16 It will be interesting to examine whether rearrangements of the AF3p21gene occur in those malignancies.