In the pediatric cancer alveolar rhabdomyosarcoma, characteristic t(2;13)(q35;q14) or variant t(1,13)(p36;q14) chromosomal translocations generate PAX3-FKHR or PAX7-FKHR fusion genes. Using fluorescence in situ hybridization, reverse transcriptase-polymerase chain reaction and quantitative Southern blot analyses, we demonstrate that these fusion genes are amplified in 20% of fusion-positive tumors. In particular, we found in vivo amplification of these fusions in one of 22 PAX3-FKHR-positive cases and five of seven PAX7-FKHR-positive cases. These findings indicate that translocation and amplification can occur sequentially in a cancer to alter both the structure and copy number of a gene and thereby activate oncogenic activity by complementary mechanisms.
The identification of nonrandom chromosomal alterations associated with specific neoplasms has resulted in the elucidation of important steps in tumorigenesis and new strategies for tumor diagnosis and management. Two important cytogenetic categories identified by these investigations are translocation and amplification. Molecular biology studies of translocations in leukemias and solid tumors have shown that these events activate oncogenes by one of two mechanisms (1). One class of translocations, which frequently occurs in lymphoid malignancies, juxtaposes the coding region of one gene into the vicinity of expression elements of another gene, resulting in increased expression and/or altered regulation. In the second category of translocations, a chimeric gene product is formed by the joining of two coding regions. For amplification events, additional copies of a cellular gene are generated in the form of double minute chromosomes or homogeneously staining regions (2). The presence of additional gene copies generally results in overproduction of wild-type transcripts and associated protein products.
Cytogenetic investigations of the pediatric soft tissue cancer alveolar rhabdomyosarcoma have identified the frequent presence of chromosomal translocations and amplification (3). In a review of 28 published cases (4), a characteristic t(2,13)(q35;q14) translocation and a variant t(1,13)(p36;q14) translocation were identified in 64 and 18% of the cases respectively. Subsequent molecular biology studies have demonstrated that these translocations fuse the PAX3 gene on chromosome 2 or the PAX7 gene on chromosome 1 with the FKHR gene on chromosome 13 to generate PAX3-FKHR or PAX7-FKHR fusion genes (5–8). These genes encode chimeric transcription factors which, in the case of PAX3-FKHR, have been shown to excessively activate transcription from binding targets of the wild-type PAX3 transcription factor (9). Molecular assays for these fusion genes have identified additional cases with cytogenetically undetectable fusions so that the overall frequency of PAX3/PAX7-FKHR fusions in alveolar rhabdomyosarcoma is >90% (10, 11, Barr et al., unpublished data).
Cytogenetic evidence of gene amplification was detected in eight of the 28 published cases (29%) of alveolar rhabdomyosarcoma (4). Two studies have used quantitative Southern blot assays to demonstrate amplification of the MYCN gene in four of six cases and three of seven cases of alveolar rhabdomyosarcoma (12, 13). The wild-type MYCN protein product functions as a transcription factor regulating expression of growth-related genes and has previously been shown to be overproduced in neuroblas-toma as a result of MYCN gene amplification (14). The amplification of other genomic loci in alveolar rhabdomyosarcoma is indicated by the finding of one case with amplification of the 12q region containing the GLI and CDK4 genes (15) and several cases with double minute chromosomes without detectable amplification of MYCN or GLI (16).
In the current study, we present FISH, RT-PCR and quantitative Southern blot experiments that demonstrate in vivo amplification of the PAX3-FKHR or PAX7-FKHR fusion gene in 20% of fusion-positive alveolar rhabdomyosarcoma cases. These findings indicate the sequential occurrence of chromosomal translocation and amplification events in a cancer, resulting in alteration of both the structure and copy number of a genomic locus. Furthermore, these findings lend further support to the premise that the 5′ PAX3-3′ FKHR and 5′ PAX7-3′ FKHR genomic fusions represent the functionally important consequences of the t(2;l3) and t(1, 13) translocations.
Molecular cytogenetic detection of amplification involving PAX3 and/or FKHR
We have previously developed a FISH assay that detects fusion of the PAX3 and FKHR loci in interphase alveolar rhabdomyosarcoma cells (17). In the course of screening rhabdomyosarcoma specimens, we identified four cases in which there were numerous hybridization signals corresponding to the PAX3 and/or FKHR loci. In case 9011, the presence of a PAX3-FKHR fusion was indicated by the presence of overlapping red and green signals (Fig. 1A). This fusion signal was present in numerous copies (>20 per nucleus) scattered throughout the interphase nucleus. In cases 319, 1340 and 1450, no fusion signals were detected. While the green signal corresponding to the PAX3 locus was present in two to four copies in these tumors, the red signal corresponding to the FKHR locus was present in numerous copies distributed throughout the nucleus (data not shown). These numerous widely-scattered hybridization signals in cases 319, 1340, 1450 and 9011 aie consistent with the presence of double minute chromosomes.
Standard cytogenetic analysis was also performed on these cases at the time of tumor diagnosis. An informative tumor karyotype was obtained only on case 319 and revealed a tetraploid karyotype with double minute chromosomes (data not shown). In addition, no t(2, 13) or t(1;13) chromosomal translocation was detected in this tumor. Therefore, in one case, both karyotypic and FISH studies indicate the presence of amplification.
RT-PCR detection of PAX3-FKHR and PAX7-FKHR fusions
These four cases were then analyzed for PAX3-FKHR and PAX7-FKHR chimeric transcripts. We developed a consensus 5′ PAX3/PAX7 primer and 3′ FKHR primer which efficiently amplify either chimeric transcript in an RT-PCR assay of tumor RNA. Using this assay, chimeric transcripts were detected in all four cases (Fig. 2A). To distinguish between PAX3-FKHR and PAX7-FKHR fusions, we hybridized labeled PAX3-and PAX7-specific oligo-nucleotide probes to the PCR products. This hybridization experiment revealed that case 9011 expressed a PAX3-FKHR chimeric transcript whereas cases 319, 1340 and 1450 expressed a PAX7-FKHR chimeric transcript. These results are consistent with the FISH results which only demonstrated a PAX3-FKHR fusion in case 9011. The presence of a PAX7-FKHR fusion in case 319 and absence of a detectable t(1;13) translocation in the corresponding karyotype indicates the presence of a cryptic translocation.
Molecular cytogenetic detection of PAX7-FKHR fusion and amplification
To determine whether the multiple FKHR hybridization signals in cases 319, 1340 and 1450 represent wild-type FKHR or PAX7-FKHR fusions, we developed a FISH assay that detects fusion of the PAX7 and FKHR loci in interphase cells. In this assay, we employed a bacteriophage clone containing the 5′ PAX7 region (18) in combination with the 3′ FKHR cosmid probe (17). Hybridization of these two labeled probes to cases 319, 1340 and 1450 revealed the presence of overlapping red and green signals indicative of PAX7-FKHR fusions (Fig. 1B). Furthermore, these fusion signals were present in multiple copies (from 7 to >20 per nucleus) scattered throughout the interphase nucleus. Therefore, the combination of FISH, cytogenetic and RT-PCR data indicate that the PAX7-FKHR fusion is amplified in cases 319, 1340 and 1450 whereas the PAX3-FKHR fusion is amplified in case 9011.
Southern blot detection of PAX3-or PAX7-FKHR amplification and rearrangement
To confirm and extend these findings of PAX3-FKHR and PAX7-FKHR amplification, we used a quantitative Southern blot assay to analyze the relative copy number of FKHR, PAX3 and PAX7 sequences which flank the t(2, 13) and t(1, 13) breakpoint regions. Genomic DNA was available from cases 319, 1450 and 9011. For the FKHR gene, probes were isolated from the 5′ region of the first intron and the third exon and restriction endonucleases were selected which did not cross the translocation breakpoints and thus only generated wild-type restriction fragments. The probes were simultaneously hybridized to restriction endonuclease digests of genomic DNA from cases 319, 1450 and 9011 and compared with genomic DNA isolated from a cytogenetically normal lymphoblast cell line (Fig. 3A and B). Visual examination of the autoradiograms demonstrates relatively comparable 5′ FKHR intron 1 hybridization signals in the tumor and lymphoblast DNA. In contrast, the hybridization signals corresponding to the FKHR exon 3 probe are substantially higher in the three tumors than the corresponding signal in lymphoblast DNA. After quantifying the hybridization signals by phosphorimaging, the ratio of the exon 3 signal to the 5′ intron 1 signal in cases 319, 1450 and 9011 is calculated to be 5-, 8-and 13-fold higher respectively, than this ratio in normal lymphoblasts. Therefore, there is an increased number of copies of the 3′ end of the FKHR gene in these tumors. The calculated ratios are likely to be underestimates of the copy number due to the probable presence of normal stromal and lymphoid cells within the tumor samples.
To further distinguish between the alternatives of aneuploidy and amplification, the blots were also hybridized with a probe for locus D13S27. This locus is situated proximal to FKHR in chromosomal region 13pter-q12 (19, 20) and is located 3′ with respect to the orientation of the FKHR gene (5, 7). The ratio of the D13S27 hybridization signal to the 5′ FKHR intron 1 signal in cases 319, 1450 and 9011 is comparable to the signal ratio in control lymphoblast DNA (data not shown). Therefore, the D13S27 locus is not present in increased copy number, indicating that only a small chromosomal segment containing the 3′ FKHR region is amplified in the three tumors.
Similar Southern blot analyses were then performed with hybridization probes from the PAX3 and PAX7 genes. Comparison of the hybridization signals of a PAX3 exon 6/7 probe and an intron 7 probe demonstrated an increased copy number of the 5′ region of PAX3 relative to the 3′ region in case 9011 (Fig. 3C). In contrast, the 5′ and 3′ PAX3 regions are present in comparable levels in cases 319 and 1450 (data not shown). However, when 5′ and 3′ PAX7 probes are hybridized to these blots, the autoradio-gram clearly demonstrates an increased relative copy number of the 5′ PAX7 region in cases 319 and 1450 (Fig. 3D) and no increased copy number in case 9011 (data not shown). These experiments therefore demonstrate that the 5′ PAX3 and 3′ FKHR regions are amplified in case 9011 whereas the 5′ PAX7 and 3′ FKHR regions are amplified in cases 319 and 1450.
We next selected restriction endonucleases whose recognition sites flank the translocation breakpoint in order to identify the rearranged genomic fragment and specifically demonstrate that the rearrangement is amplified. In particular, we focused on the PAX3 rearrangement in case 9011 because of our extensive mapping information and the relatively small size (∼18 kb) of
PAX3 intron 7, in which the t(2, 13) breakpoints occur (5, 21). Hybridization of the 5′ PAX3 probe (exon 6/7) to PstI, NcoI and ApaLI digests of case 9011 revealed novel fragments which represent the 5′ PAX3-3′ FKHR genomic fusion (Fig. 4A). In all cases, the hybridization intensity of the rearranged signal was substantially greater than that of the unrearranged wild-type signal in the tumor or the lymphoblast control DNA. In contrast, the 3′ PAX3 probe (intron 7) hybridized to novel HpaI and MspI fragments in case 9011 which correspond to the reciprocal 5′ FKHR-3′ PAX3 genomic fusion (Fig. 4B). These reciprocal rearranged fragments are not present in increased copy number relative to the wild-type restriction fragment in the tumor or lymphoblast control DNA. These findings demonstrate that the 5′ PAX3-3′ FKHR genomic fusion is specifically amplified in case 9011 whereas the reciprocal 5′ FKHR-3′ PAX3 fusion and wild-type PAX3 gene are present but not amplified.
Prevalence of amplification in alveolar rhabdomyosarcomas
We then used the assays described above to screen a large series of alveolar rhabdomyosarcomas and determine the prevalence of amplification of the PAX3-FKHR or PAX7-FKHR fusion genes. Using the consensus RT-PCR assay, we identified 29 cases expressing the PAX3-FKHR or PAX7-FKHR fusion which also had material available for the Southern blot or FISH assays (Table 1). The FISH assay with the 5′ PAX3 and 3′ FKHR cosmid probes and/or the quantitative Southern blot assay with the flanking FKHR probes were performed on these cases. We set the minimum criteria for amplification as six hybridization signals or a normalized 3′ FKHR to 5′ FKHR ratio of four. Of 22 cases expressing a PAX3-FKHR chimeric transcript, case 9011 was the only one with evidence of amplification. In contrast, five of seven cases expressing a PAX7-FKHR transcript (including cases 319, 1340 and 1450) demonstrated evidence of PAX7-FKHR amplification. In two additional tumors (cases 520 and 1181) detected by the Southern blot assay, the normalized 3′ FKHR to 5′ FKHR ratio was calculated to be eight (Fig. 5). Therefore, amplification of the fusion gene was present in six of 29 cases of fusion-positive alveolar rhabdomyosarcoma and the majority of these amplification events involved the variant PAX7-FKHR fusion which arises from the t(1;13) translocation.
To determine the relationship between amplification of these fusions genes and amplification of the MYCN gene in alveolar rhabdomyosarcomas, the MYCN copy number was assessed in the tumor cells by two approaches. The number of MYCN copies was directly visualized in tumor cells by FISH analysis with a MYCN probe. By hybridization of Southern blots of tumor DNA with probes for MYCN and a control chromosome 2 locus (HOX4D located at 2q31), we also determined the ratio of the MYCN hybridization signal to that of the control locus and normalized this number for the signal ratio in a normal lymphoblast cell line. Using the criteria for amplification discussed above, we detected MYCN amplification in three of 27 tumors (cases 20, 1190 and 1477) (Table 1). Amplification of MYCN was not detected in the six tumors with amplification of the PAX3-FKHR or PAX7-FKHR fusion genes and thus there is no evidence of a propensity for different loci to be amplified within the same tumor.
Previous studies have demonstrated that, in most alveolar rhabdo-myosarcomas, a chromosomal translocation event juxtaposes PAX3 or PAX7 with FKHR and thereby generates fusion genes which encode chimeric protein products (5–9). In this report, we have employed FISH, RT-PCR and quantitative Southern blot approaches to demonstrate that, in a subset of fusion-positive alveolar rhabdo-myosarcomas, a second event occurs which results in amplification of the PAX3-FKHR or PAX7-FKHR fusion genes. Therefore, two sequential chromosomal alterations have occurred during the development of these tumors, first affecting the structure and coding information of specific genomic loci and then affecting the copy number of these structurally altered loci. These findings demonstrate a novel situation in which translocation and amplification events cooperate to generate a potent oncogenic activity.
A striking result of our experiments is the finding that the PAX7-FKHR fusion gene was amplified in five of seven PAX7-FKHR-positive cases whereas amplification of the PAX3-FKHR fusion was only detected in one of 22 PAX3-FKHR-positive cases. Several potential explanations for the higher frequency of cases with PAX7-FKHR amplification can be proposed. The possibility that the PAX7-FKHR product stimulates gene amplification is unlikely because the DNA binding domains of PAX3 and PAX7 are highly homologous (6, 22, 23) and the PAX3-FKHR and PAX7-FKHR fusion transcription factors are postulated to have very similar activities. A potential difference between the 5′ PAX3 and PAX7 regions is the elements regulating the expression of the fusion genes, which may correspond to the differences in developmental expression of the wild-type PAX3 and PAX7 genes (22, 23). If the PAX7 expression elements ate less powerful than the PAX3 elements, then gene amplification may be one strategy to increase the level of PAX7-FKHR transcript expression above a needed threshold for oncogenic activity.
A third explanation for the relative amplification frequencies is the orientation of the genes on their respective chromosomes. We have previously proposed, based on somatic cell hybrid experiments, that the 5′ ends of PAX3 and FKHR are each oriented towards the telomeres of chromosomes 2 and 13 (5, 7). This symmetric orientation permits a functional fusion to be formed following single breaks within each gene and reciprocal exchange of chromosomal fragments. The orientation of PAX7 on chromosome 1 is not currently known and the necessary reagents for this determination are still being developed. If the 5′ end of PAX7 is oriented towards the centromere of chromosome 1, then a simple translocation event with single breaks in PAX7 and FKHR would not produce a functional fusion. Instead, a second break outside PAX7 or FKHR followed by inversion of one of the two genes would be needed to correctly orient these genes. This second break would create a second free end that may facilitate additional recombination events (24).
Cytogenetic investigations of alveolar rhabdomyosarcoma have frequently identified both reciprocal derivative chromosomes, der(13) and either der(2) or der(1), resulting from the t(2,13) or t(1,13) translocations (3). Our previous RT-PCR investigation of the chimeric transcripts associated with the t(2, 13) translocation showed that the 5′ PAX3-3′ FKHR and 5′ FKHR-3′ PAX3 transcripts are detectably expressed in many cases from the der(13) and der(2) respectively (7). However, other experiments have revealed several important differences between these reciprocal chimeric transcripts. The 5′ PAX3-3′ FKHR transcript is much more highly and consistently expressed (5, 7) and has been shown to encode a functional transcription factor with an intact PAX3 DNA binding domain and intact FKHR activation domain (9). The experiments in the current report demonstrate that when amplification occurs, this process preferentially involves the 5′ PAX3-3′ FKHR and 5′ PAX7-3′ FKHR fusion genes. Therefore, these pieces of evidence consistently indicate that 5′ PAX3-3′ FKHR and 5′ PAX7-3′ FKHR represent the functionally important fusions involved in the pathogenesis of alveolar rhabdomyosarcoma.
The propensity for amplification of the 5′ PAX3-3′ FKHR and 5′ PAX7-3′ FKHR fusion genes is another important consideration in understanding the cytogenetics of alveolar rhabdomyo-sarcoma. Several cases without a detectable der(13) have been previously reported (25, 26). In fact, the visible presence of the der(2) and absence of an identifiable der(13) in some cases initially suggested that the der(2) encoded the important functional product. RT-PCR and FISH analyses have subsequently shown that the fusion genes corresponding to the der(13) are consistently present and expressed (7, 8). Such discordant findings were previously explained by the difficulty in definitively identifying the small der(13) or the occurrence of more complex alterations which rearranged the der(13) with other chromosomes. Our current findings indicate that an additional important possibility to explain the absence of an identifiable der(13) is the occurrence of amplification and thus the presence of the fusion genes on extra-chromosomal elements.
Materials And Methods
Standard and molecular cytogenetics
Tumor samples were collected and characterized histopathologi-cally as described (10). Portions of fresh tumor samples were dissociated with sterile scalpel blades in RPMI supplemented with 15% fetal bovine serum and antibiotics and processed for cytogenetic studies as described (27). Cell pellets from direct preparations or 24 h cultures were used for the FISH assays. Cosmids containing the 5′ PAX3 region (CWSI22, provided by Y. Nakamura) (28) and the 3′ FKHR (clone 5) (17) region were labeled with biotin-14-dATP and digoxigenin-11-dUTP respectively. Clones containing the 5′ PAX7 region (HuP1 phage, provided by M. Noll) (18) and the 3′ FKHR region (cosmid clone 5) (17) were labeled with digoxigenin-11-dUTP and biotin-14-dATP respectively. After hybridizing these labeled probes on slides, detection of biotin and digoxigenin was accomplished with FITC-avidin and anti-digoxigenin rhodamine, as described previously (17). A probe for the MYCN locus (Oncor) was hybridized and detected according to the manufacturer's recommendations.
Reverse transcriptase-polymerase chain reaction assays
Total RNA was isolated, treated with DNase I and reverse transcribed. PCR amplification of the cDNA product was performed as previously described (6). The primers for the PAX3/PAX7-FKHR consensus assays are 5′ PAX3/PAX7 (CCGACAGCAGCTCTGCCTAC) and 3′ FKHR (10). To verify the presence of intact RNA and amplifiable cDNA, each reverse transcription reaction product was also assayed for wild-type FKHR expression with 5′ FKHR and 3′ FKHR primers (10).
PCR reaction products (10 µl) were electrophoresed on 2% agarose gels and stained with ethidium bromide. For slot blotting, aliquots of the PCR products were denatured and applied to a nylon membrane using a filtration manifold. PAX3-and PAX7-specific oligonucleotides are CTACTGCCTCCCCAG-CAC and CACAGCTTCTCCAGCTACTCTG respectively.
These oligonucleotides were end-labeled with [γ-32P]ATP and 10 U T4 polynucleotide kinase, hybridized to the slot blot filters at 42°C and washed in 0.5 × SSC, 0.1% SDS at 42°C.
Genomic DNA analysis
Genomic DNA was isolated, digested with restriction endo-nucleases, electrophoresed and blotted to nylon membranes, as described previously (20). The hybridization probes are listed in Table 2 and were isolated, labeled and hybridized to membranes as described previously (20). Intensities of bands on the hybridized membrane were quantified using a phosphorimager (Molecular Dynamics).
The authors wish to thank G. Brodeur and P. White for their advice. We acknowledge the excellent technical assistance provided by Q.-B. Xiong and J. Hollows. We also appreciate the assistance of the Children's Cancer Group and the cooperation of the institutional oncologists and pathologists who provided specimens. This work was supported by funds from NIH Grants CA64202 to F.G.B. and J.A.B., CA61935 to F.G.B. and CA13539 to the Children's Cancer Group and ACS Grant JFRA-492 to F.G.B.
fluorescence in situ hybridization
reverse transcriptase-polymerase chain reaction
derivative chromosome 1
derivative chromosome 2
derivative chromosome 13