Genetic Heterogeneity in the Alveolar Rhabdomyosarcoma Subset without Typical Gene Fusions

INTRODUCTION

RMS 3 is a family of soft tissue tumors related to the skeletal muscle lineage that occur in children and young adults (1) . On the basis of histopathologic appearance, RMS has been divided into two major subtypes, ARMS and ERMS. Cytogenetic analyses have revealed recurrent t(2;13)(q35;q14) and t(1;13)(p36;q14) chromosomal translocations in ARMS but not in ERMS cases. Molecular genetic studies demonstrated that these translocations disrupt chromosomes 2 and 1 within the PAX3 and PAX7 loci, respectively, that encode related members of the paired box family of transcription factors (2 , 3) . These genes are juxtaposed with portions of the FKHR gene on chromosome 13, which encodes a member of the fork head family of transcription factors (3 , 4) . The end result is the formation of chimeric genes that are expressed as chimeric transcripts, which can be readily detected with RT-PCR methodology (5 , 6) . These novel transcripts are in turn translated into chimeric proteins in which the DNA binding domains of PAX3 or PAX7 are joined to the transcriptional activation domain of FKHR to generate novel transcription factors with oncogenic activity (7, 8, 9, 10) .

In a clinical correlative study of ARMS cases from IRS-IV, differences were detected between cases with PAX3-FKHR and PAX7-FKHR fusion transcripts (11) . In particular, PAX7-FKHR-expressing tumors occurred in younger children, were locally less invasive, and yet showed a comparable frequency of metastasis compared with PAX3-FKHR-expressing tumors. Although there was no significant survival difference between the two fusions in patients with locoregional tumors, PAX7-FKHR was associated with a significantly better outcome than PAX3-FKHR in patients with metastatic disease. This difference in outcome may be related to the higher propensity of PAX3-FKHR-expressing tumors to metastasize to bone marrow.

In the above-described IRS-IV study (11) as well as in previous molecular diagnostic studies of ARMS (5 , 12, 13, 14, 15, 16) , PAX3-FKHR and PAX7-FKHR transcripts were detected in ∼80% of ARMS cases, and thus there is a fusion-negative subset comprising ∼20% of ARMS cases. In the IRS-IV study, these fusion-negative cases had clinical characteristics that were generally intermediate between the PAX3-FKHR and PAX7-FKHR categories without statistically significant differences with either category (11) . On the basis of the lack of a distinctive clinical phenotype, we hypothesized that this fusion-negative category may be genetically heterogeneous. To address this hypothesis, we investigated whether this category may contain cases with alterations of the PAX3, PAX7, or FKHR loci that were not detectable with the standard RT-PCR assays. Using RT-PCR, FISH, and Southern blot methodologies, we identified several distinct genetic categories within this “fusion-negative” subset.

MATERIALS AND METHODS

Tumor Specimens.

Frozen tumor samples were retrieved from the Intergroup Rhabdomyosarcoma Study Group/Pediatric Cooperative Human Tissue Network tumor bank in Columbus, Ohio, as well as institutional tumor banks of the participating investigators. All of the cases were reviewed at the Intergroup Rhabdomyosarcoma Study Group Biopathology Center (Columbus, OH), and histopathological diagnoses were based on the International Classification of Rhabdomyosarcoma criteria (17) . All of the ARMS cases showed either the classic cystic or solid alveolar growth patterns. Dependent on the degree of intervening stroma, cells grew in nests or cords/trabeculae with either nascent (micro-alveolar) or frank central cystic change. Characteristic nuclear features of ARMS were present with monomorphic cells showing nuclei with coarse chromatin and prominent nucleoli or evenly distributed nuclear chromatin. Tumor giant cells were also present in a subset of these cases. Finally, in all of the cases, the majority of tumor cells were viable, as assessed by histological analysis.

RNA Extraction and RT-PCR Analysis.

Primers and probes are listed in Table 1 ⇓ and illustrated in Fig. 1 ⇓ . Total RNA was extracted from frozen primary tumors using the acid-guanidinium-phenol-chloroform method as described previously (6) . In the standard sensitivity consensus RT-PCR assay for the PAX3-FKHR and PAX7-FKHR fusion transcripts, RNA was pretreated with DNase I, reverse transcribed from random hexamers, and the cDNA was amplified with PAX3/PAX7-1 and FKHR-R primers, as described previously (6) . In the high sensitivity consensus RT-PCR assay for PAX3-FKHR and PAX7-FKHR transcripts, DNase I-treated RNA was reverse transcribed with the FKHR-RT2 primer, and the cDNA was amplified with the PAX3/PAX7-1 and FKHR-2 primers, as described previously (18) . In an alternative high sensitivity assay, DNase I-treated RNA was reverse transcribed with the FKHR-RT1 primer, and the cDNA was amplified with the PAX3/PAX7-1 and FKHR-R primers (19) . Each cDNA preparation was assayed for wild-type FKHR expression with FKHR-4 and FKHR-R (standard sensitivity) or FKHR-4 and FKHR-2 (high sensitivity) primers as a control for the presence of intact RNA.

PCR products were fractionated by agarose gel electrophoresis, stained with ethidium bromide, and then transferred to nylon filters. The blots were hybridized with ³²P-labeled oligonucleotide probes: FKHR-P (standard sensitivity) and PAX3/PAX7-FKHR or FKHR-P1 (high sensitivity). To distinguish fusion products involving PAX3 and PAX7, PCR products were denatured and applied to nylon membranes with a slot blot filtration manifold, as described previously (6) . These blots were then hybridized with ³²P-labeled oligonucleotide probes PAX3-P and PAX7-P.

In RT-PCR assays for PAX3/PAX7-AFX and PAX3/PAX7-FKHRL1 fusion transcripts, DNase I-treated RNA was reverse transcribed with random hexamers. In a consensus assay for PAX3/PAX7-FKHR and PAX3/PAX7-AFX fusion transcripts, the cDNA was amplified with PAX3/PAX7-1 and FKHR-2 primers, and the PAX3/PAX7-AFX PCR products on blots were detected with the AFX-R3 probe. In a consensus assay for PAX3/PAX7-FKHR and PAX3/PAX7-FKHRL1 fusion transcripts, the cDNA was amplified with PAX3/PAX7-1 and FKHRL1-R1 primers, and the PAX3/PAX7-FKHRL1 PCR products on blots were detected with the FKHRL1-R2 probe. In a specific assay for PAX3/PAX7-AFX fusion transcripts, the cDNA was amplified with PAX3/PAX7-1 and AFX-R1 primers, and the PCR products on blots were hybridized with the AFX-R2 probe. All of the PCR reactions and hybridizations were performed with conditions described previously (6 , 18 , 19) .

DNA Extraction and Southern Blot Analysis.

Genomic DNA was isolated from frozen specimens by the Qiagen Genomic DNA Purification Procedure. For detection of PAX3 rearrangements, DNA aliquots (2.5–5 μg) were digested with HindIII or PstI, fractionated on 0.75% agarose gels, blotted to nylon membranes, and sequentially hybridized with gel-purified PAX3 probes labeled by random priming (20) . For detection of PAX7 rearrangements, DNA aliquots (2.5–5 μg) were digested with HindIII or ApaLI, and similarly electrophoresed, blotted, and sequentially hybridized with the PAX7 probes (21) .

FISH Analysis.

Touch imprints on glass slides were prepared from frozen tumor specimens using standard techniques. Rearrangements of the FKHR gene were detected by a splitting assay with cosmids FKHR-1 and FKHR-5 containing the 5′ and 3′ regions of the FKHR locus, respectively (22) . PAX3-FKHR and PAX7-FKHR gene fusions were detected by a fusion assay with a P1-derived artificial chromosome clone containing the FKHR genomic region (930-E10) in combination with a P1-derived artificial chromosome clone containing either the PAX3 (1126-C17) or PAX7 (394-P21) genomic region. Probes were differentially labeled using the Biotin-Nick Translation mixture (Boehringer Mannheim) and DIG-Nick Translation mixture (Roche), following the manufacturer’s labeling protocols. Labeled probes were mixed with Cot 1-DNA, precipitated in ethanol, resuspended in Hybrisol VII (Oncor, 50% formamide/2× SSC), denatured for 5 min at 73°C, and preannealed at 37°C for 1 h. Slides containing interphase cells from frozen tumor specimens were treated sequentially with RNase and pepsin, fixed with formaldehyde, denatured, dehydrated with ethanol, and then dried, hybridized with denatured probe, and washed, as described previously (23) . Probe detection was performed with FITC-avidin and antidigoxigenin rhodamine (Ventana), and 4′,6-diamidino-2-phenylindole was used as nuclear counterstain. Slides were examined using a Zeiss Axioplan epifluorescence microscope equipped with a COHU CCD camera. Three hundred cells were evaluated per sample. Images were captured using PSI Powergene software and then processed using Adobe Photoshop 6.0. Positive controls included two PAX3-FKHR-positive ARMS cases, whereas negative controls included ERMS and neuroblastoma cases.

RESULTS

Identification of Fusion-negative ARMS Cases.

The basic criteria for inclusion in this study and categorization as a fusion-negative ARMS are: (a) a review diagnosis of ARMS;(b) availability of frozen tumor tissue or isolated tumor RNA; (c) absence of a PAX3-FKHR or PAX7-FKHR transcript in a standard sensitivity RT-PCR assay; and (d) presence of a wild-type FKHR transcript in a standard sensitivity RT-PCR assay. To ensure the consistency of our molecular testing, all of the putative fusion-negative ARMS cases and selected fusion-positive cases from the participating laboratories were re-evaluated by one reference laboratory (F. G. B.; Fig. 2A ⇓ ). The adequacy of an RNA sample was evaluated by RT-PCR analysis of wild-type FKHR expression. Because wild-type FKHR mRNA is expressed at comparable or lower levels than PAX3-FKHR and PAX7-FKHR mRNA in ARMS tumors (24) , the detection of a wild-type FKHR PCR product assures that the tumor RNA is suitably intact and that there is adequate sensitivity to detect fusion products. The PAX3/PAX7-FKHR fusion transcript status of these cases was assessed with a standard sensitivity consensus RT-PCR assay (described in “Materials and Methods”; Ref. 6 ). In addition to evaluating the presence of fusion PCR products on the ethidium bromide-stained agarose gel, all of the PCR products from the consensus PAX3/PAX7-FKHR assay were transferred to nylon membranes and hybridized with an FKHR-specific oligonucleotide probe. Although faint hybridizing signals were seen in a few cases (such as sample 2 in Fig. 2A ⇓ ), these findings were not reproducible among the several independent runs. Therefore, based on a requirement of at least two positive results to score a case as fusion-positive and the requirement of at least two completely negative results to score a case as fusion-negative, we identified 23 fusion-negative ARMS cases (from a collection of ∼120 ARMS cases) for additional evaluation (Table 2) ⇓ . Additional analysis of many of these cases also demonstrated the absence of gene fusions characteristic of other sarcoma categories (5 , 6) , such as the EWS-FLI1 and EWS-ERG fusions associated with the Ewing family of tumors (data not shown).

Application of High Sensitivity RT-PCR Assays.

One explanation for our finding of a few cases with irreproducible hybridizing signals in the PAX3/PAX7-FKHR assay was the possibility of very low expression of a fusion transcript. To evaluate this possibility, we applied an RT-PCR assay originally designed for detection of these fusion transcripts in RNA derived from paraffin blocks (18) . In this assay, the sensitivity of fusion product detection was increased by use of an FKHR-specific reverse transcription primer, decreasing the size of the target PCR product, and increasing the PCR cycle number. On the basis of studies of serial dilutions of ARMS RNA with lymphoid RNA, we estimate that such modifications increase the assay sensitivity by one to two orders of magnitude. Furthermore, the near-maximal optimization of the PCR component of this and other reactions is indicated by the finding that assay sensitivity is not improved by a nested PCR protocol (19) . Application of this high sensitivity PAX3/PAX7-FKHR RT-PCR assay to the 23 fusion-negative ARMS cases demonstrated consistent detection of a fusion PCR product in four samples (cases 1, 2, 14, and 16; Fig. 2B ⇓ ). It should be noted that, in all four of the cases, the fusion product from the high-sensitivity assay was evident on the ethidium bromide-stained agarose gel. As a test of the specificity of this assay, we analyzed 20 ERMS cases that were shown previously to be fusion-negative with the standard sensitivity assay and did not detect any fusion products with the high-sensitivity assay (data not shown).

To determine whether the PCR products detected by this high sensitivity assay represent low expression of standard PAX3-FKHR or PAX7-FKHR fusions, or expression of an aberrant species, we first hybridized a Southern blot with a PAX3/PAX7-FKHR junctional probe (Fig. 2, B and C) ⇓ . The PCR products from cases 1, 2, and 14 demonstrated strong hybridization to the junctional probe, whereas the PCR product from case 16 consistently demonstrated weaker hybridization suggesting that it is not a PAX3-FKHR or PAX7-FKHR fusion. We then directly sequenced the three strongly hybridizing PCR products, and identified a PAX3-FKHR fusion in one case (case 14) and a PAX7-FKHR fusion in the other two cases (cases 1 and 2). On the basis of these studies, we propose that these cases express low levels of the PAX3-FKHR or PAX7-FKHR fusions, which cannot be reliably detected by the standard-sensitivity RT-PCR assay.

Additional controls were performed to further verify that these cases represent low level fusion transcript expression. To rule out the possibility of contaminating PCR products resulting in false-positive results, these assays were repeated without reverse transcriptase enzyme, and no evidence of such contaminating species was evident on either the stained gel or the autoradiogram after hybridizing a Southern blot with a fusion junction-specific probe (data not shown). When an alternative high-sensitivity assay that uses the same PCR primers as the standard sensitivity assay was used (19) , these three cases similarly generated the expected larger products (data not shown), indicating that the fusions in these three cases are not structurally different from those detectable by the standard assay.

Identification of a PAX3-AFX Gene Fusion.

In case 16, which was positive on the high-sensitivity assay but did not demonstrate strong hybridization to the PAX3/PAX7-FKHR junctional probe, sequencing of the PCR product revealed a fusion of PAX3 to sequences from AFX, a member of the fork head family closely related to FKHR and localized to chromosomal region Xq13 (25) . Because of the high sequence similarity between FKHR and AFX in the vicinity of the reverse FKHR primer (FKHR-2; Fig. 1 ⇓ ), this high-sensitivity assay was fortuitously able to bind to and amplify the PAX3-AFX fusion transcript. In contrast, there was significantly less sequence similarity in the vicinity of the FKHR-R primer and, thus, the standard sensitivity assay would not be expected to amplify this alternative fusion transcript. To verify the nature of this fusion species, an AFX-specific oligonucleotide probe was designed from a region in which AFX and FKHR sequences are dissimilar. This probe demonstrated strong hybridization to the putative PAX3-AFX amplification product, and no hybridization to the PAX3-FKHR and PAX7-FKHR products from the other three cases (Fig. 2C) ⇓ . Furthermore, we designed an AFX reverse primer to use in conjunction with the PAX3/7-1 forward primer and used this PAX3/PAX7-AFX-specific RT-PCR assay to amplify the PAX3-AFX fusion from case 16 but not a PAX3-FKHR-expresssing control (Fig. 2D) ⇓ .

Within the fork head family, FKHR, AFX, and FKHRL1 constitute a subfamily with highly similar sequence in the fork head domain and additional regions of sequence similarity outside the fork head domain (26) . To investigate the possibility of fusions of PAX3 or PAX7 to AFX or FKHRL1, we designed additional consensus RT-PCR assays that would detect these novel fusions in addition to the standard PAX3/PAX7-FKHR fusions (Fig. 1) ⇓ . The high-sensitivity PAX3/PAX7-FKHR assay described previously was adapted to a PAX3/PAX7-FKHR/AFX consensus assay by substituting random hexamers for the FKHR-specific reverse transcription primer. In addition, a reverse primer (FKHRL1-R1) was identified from a region of sequence similarity between FKHR and FKHRL1 to develop a PAX3/PAX7-FKHR/FKHRL1 consensus assay. Analysis of the other fusion-negative cases did not reveal any evidence of these novel fusions, either on ethidium bromide-stained gels or after hybridization of Southern blots with AFX- or FKHRL1-specific oligonucleotide probes (data not shown).

Southern Blot Analysis of PAX3 and PAX7 Rearrangements.

Our previous analysis of fusion-positive ARMS cases demonstrated that the t(2;13) and t(1;13) breakpoints consistently disrupt the seventh introns of the PAX3 and PAX7 loci, which span 17.5 kb and 32 kb, respectively (20 , 21) . These breakpoints are situated to maintain the integrity of the paired box and homeobox domains, and separate them from an essential part of the transcriptional activation domain. To screen our fusion-negative ARMS cases for rearrangements that fuse PAX3 or PAX7 to novel loci, we used a Southern blot approach that assays genomic DNA for rearrangements of PAX3 and PAX7 introns 7 and 8 (and part of intron 6; Fig. 3A ⇓ ). Because of the functional organization of these genes, breakpoints in more 5′ regions would exclude the homeodomain and not generate oncogenic fusion proteins, and, thus, these 5′ regions were not examined in this study. By hybridizing a series of probes to digests of genomic DNA with two restriction enzymes, we designed assays that would assess the structural integrity of the 3′regions. Although intron 6 is not fully screened by this Southern blot analysis, it comprises only 5% of the PAX3 or PAX7 genomic regions between exons 6 and 9, and, thus, our procedure examines the vast majority of the involved intronic regions while minimizing the amount of genomic DNA required. To test the sensitivity of these assays, we analyzed a series of fusion-positive ARMS cases and detected PAX3 rearrangements in 18 of 21 PAX3-FKHR-expressing cases and PAX7 rearrangements in 8 of 9 PAX7-FKHR-expressing cases (data not shown). The few false-negatives are attributed to situations where the rearranged fragment is similar in size to the wild-type fragment or to rearrangements occurring within the few small regions not fully detected by our Southern blot strategy.

Satisfactory quality genomic DNA was available from 16 of the ARMS cases in this study for analysis of PAX3 and PAX7 rearrangements. A PAX3 rearrangement was detected in the above-described case 16 with the PAX3-AFX gene fusion (Fig. 3B) ⇓ . In two cases with low level fusion transcript expression, one with PAX3-FKHR (case 14) and the other with PAX7-FKHR (case 2), gene rearrangements were not detected, consistent with the number of cells with a PAX3 or PAX7 rearrangement being below the detection limits of this Southern blot assay. Finally, in 13 cases that were fusion-negative in the low- and high-sensitivity RT-PCR assays, a PAX7 rearrangement was detected in case 21 (Fig. 3C) ⇓ , and a PAX3 rearrangement was detected in case 24 (Fig. 3D) ⇓ . These rearrangements were verified by Southern blot assays with additional restriction enzymes, thus indicating that the novel fragments were not simply the result of restriction fragment length polymorphisms (Fig. 3D) ⇓ . Therefore, novel fusions involving PAX3 or PAX7 were present in three cases, one with a PAX3-AFX fusion and two additional cases in which the rearrangement partner is unknown.

FISH Detection of FKHR Rearrangements.

On the basis of the large size (130 kb) of FKHR intron 1, which contains the t(2;13) and t(1;13) breakpoints, we selected a FISH approach to screen the fusion-negative cases for rearrangements of the FKHR locus (22) . This assay uses two differentially labeled cosmid probes that contain inserts from the 5′and 3′ regions flanking FKHR intron 1. These reagents will screen for rearrangements in the 5′ untranslated region and intron 1, the two regions in which breakpoints would generate functional oncogenic fusions. Hybridization of these probes to negative control cases demonstrates fused signals corresponding to the wild-type FKHR loci. Because of hybridization artifacts, split signals are sometimes detected in a small subset of negative control cells, depending on the tumor cell preparation. Using negative controls prepared and stored similarly as our experimental samples, this false-positive rate is <10% of the cells and, thus, constitutes the detection limits of our assay using the available material. In ARMS tumor cells with the 2;13 or 1;13 translocations, the FKHR rearrangement results in splitting of one or more sets of hybridization signals (Fig. 4A) ⇓ . Analysis of positive control cases prepared similarly as our experimental cases demonstrated splitting of the two hybridization signals in a large majority of analyzed cells.

Satisfactory quality touch preparations were available from 15 of the ARMS cases in this study to assay for rearrangements of the FKHR locus (Fig. 4, B and C) ⇓ . There were no FKHR rearrangements detected in case 16 with the PAX3-AFX fusion as well as case 21 with the novel fusion involving PAX7. Material was not available for similar studies of case 24. In addition, FKHR rearrangements were not detected (and thus present in <10% of cells) in the three cases with low expression of PAX3-FKHR or PAX7-FKHR fusion transcripts, again indicating that these fusions were present in a small subset of cells that is below the detection limits of the assay. Of the 10 assayed cases without evidence of PAX3/PAX7 rearrangements or low fusion transcript expression, we detected evidence of FKHR rearrangements in 2 of these cases (18% and 24% of cells with split signals in cases 22 and 26, respectively; Fig. 4C ⇓ ).

To additionally analyze the latter 2 cases with FKHR rearrangements, we performed dual color FISH analysis using a PAC clone containing the FKHR genomic region (930-E10) in combination with a PAC clone containing either the PAX3 (1126-C17) or PAX7 (394-P21) genomic region. In this assay, an overlapping or fused hybridization signal is indicative of a PAX3-FKHR or PAX7-FKHR gene fusion. These FISH assays demonstrated a PAX3-FKHR fusion in 21% of cells in case 26 (Fig. 4D) ⇓ and a PAX7-FKHR fusion in 20% of cells in case 22 (data not shown), corroborating the level of FKHR splitting found with the FKHR cosmid probes, as shown above. In contrast, case 21, which did not demonstrate splitting using the FKHR probes, showed only background levels of fusion signals using the PAC probes (data not shown).

DISCUSSION

One of the striking features of gene fusions in sarcomas is the consistency with which a specific fusion occurs in the corresponding sarcoma category. Several large studies with careful pathologic review and well-controlled molecular assays have emphasized the near 100% frequency of the EWS-WT1 fusion in desmoplastic small round cell tumor (27) , the SYT-SSX1 and SYT-SSX2 fusions in synovial sarcoma (28) , and the EWS-FLI1 and variant fusions in Ewing family tumors (29) . On the basis of these studies, it can be questioned whether there truly is a fusion-negative subset of Ewing sarcoma, desmoplastic small round cell tumor, or synovial sarcoma. In contrast, our recent study of 78 ARMS cases, which involved central pathologic review and molecular diagnostic analyses in three independent laboratories, clearly identified a fusion-negative subset accounting for 20% of ARMS cases (11) . In this report, we have analyzed 23 histologically confirmed ARMS cases in which the two characteristic fusions were not detected by standard RT-PCR analysis. Using optimized RT-PCR assays for low level fusion transcript detection, Southern blot analysis of PAX3 and PAX7 rearrangements, and FISH detection of FKHR rearrangements, we have determined that this “fusion-negative” ARMS category can be divided into several genetic subsets, including low expression and/or atypical presentation of standard fusions, variant fusions, and true fusion-negative cases.

Among the 23 initial cases selected, we identified three cases in which PAX3-FKHR or PAX7-FKHR fusion transcripts could only be detected with a high-sensitivity RT-PCR assay. A similar finding was reported recently in a molecular diagnostic study of Ewing family tumors in which fusion transcripts in 4 of 58 fusion-positive cases could only be detected after a high-sensitivity nested RT-PCR procedure (30) . On the basis of the reproducibility of our findings in conjunction with the negative results obtained in assays without reverse transcription enzyme and in assays of control ERMS cases, we propose that these cases are expressing fusion transcript levels that are lower than what can be detected by our standard RT-PCR assay. Furthermore, our inability to detect PAX3, PAX7, or FKHR rearrangements in these cases by Southern and FISH assays suggests that the low expression level is because of the presence of the gene fusion in only a few cells rather than generally low expression throughout the tumor cell population. The detection limits of the Southern and FISH assays are estimated to be 10% of the assayed cells, and, thus, more sensitive assays would be needed to detect these rare fusion-positive cells with these methodologies. The question then arises as to whether the presence of a small subset of fusion-positive cells indicates that the fusion occurred late during tumor progression or alternatively that the fusion was lost during tumor progression. To provide support for the latter hypothesis, one of the authors (P. H. B. S.) described previously a case of Ewing sarcoma in which the EWS-FLI1 fusion was lost in a well-differentiated recurrence after therapy (31) .

Among 16 cases for which genomic DNA was available for Southern blot analysis, we detected rearrangements of PAX3 or PAX7 in three cases. In one case, the fusion partner was determined to be AFX, which encodes a fork head protein with a structure very similar to that of FKHR (25) . In two additional cases, rearrangements were mapped in the PAX3 and PAX7 loci within the same introns involved in the typical ARMS fusions, but neither PAX3-FKHR nor PAX7-FKHR transcripts were detected by RT-PCR. Furthermore, in one of these two cases, FISH analysis confirmed that the FKHR locus was not rearranged. Therefore, we hypothesize that other genomic loci are fused to PAX3 and PAX7 in these cases to generate variant gene fusions. Because we ruled out involvement of the two most closely related fork head genes, AFX and FKHRL1, these fusions may involve more distantly related fork head genes (32) . Similarly, in Ewing family tumors, although most fusions involve the highly related ETS genes FLI1 and ERG, rare fusions have been identified with more distantly related ETS genes, such as ETV1 and E1AF (33) . An alternative possibility is that the 3′ fusion partner is not a member of the fork head family but instead is an unrelated gene that can provide the required COOH-terminal transcriptional activation domain (34) .

In our FISH analysis of 15 cases, we found two cases in which a PAX3-FKHR or PAX7-FKHR fusion was detected by FISH analysis but was not associated with a fusion transcript detectable by RT-PCR. Although the fusion signals were only found in ∼20% of the cells, the sensitivity of our RT-PCR analysis should be more than sufficient to detect fusion transcripts from these cells. Therefore, we hypothesize that either these fusion loci are not transcribed in these tumors or that the resulting fusion transcript has some feature that precludes detection by our RT-PCR assays. Of note, in both of these cases, the PAX3 and PAX7 rearrangements were not detected by Southern blot analysis of PAX3 and PAX7 intron 7, which is disrupted in all of the cases examined previously with 2;13 and 1;13 translocations. We acknowledge that the limited Southern blot screening assay used in this study only detects rearrangements in approximately 85–90% of fusion-positive cases and, thus, the rearrangements in these two cases may have been missed because of technical restrictions. However, the alternative possibility is that other regions of the PAX3 and PAX7 loci (such as cryptic exons) are involved in these rearrangements, resulting in a fusion gene with a transcription product that is not detectable by the RT-PCR assays used in this study. These issues will be analyzed in future detailed genomic studies of these cases.

Finally, in >50% of the fusion-negative ARMS cases, there is no detectable involvement of the PAX3, PAX7, or FKHR genes. These cases may contain a variant fusion or other genetic events that recapitulate the essential biological events occurring downstream of the PAX3-FKHR and PAX7-FKHR fusions. Alternatively, although these cases have a similar histological appearance as fusion-positive ARMS, this subset of tumors may be a genetically and biologically distinct entity. In the current retrospective study, because of the lack of sufficient material in some cases to perform either Southern analysis of PAX3/PAX7 rearrangements and/or FISH analysis of FKHR rearrangements, there were only four cases in which we fully ruled out any evidence of PAX3, PAX7, and/or FKHR rearrangements. The small numbers in this group of “true fusion-negative” cases do not provide sufficient statistical power for detailed clinical and pathological comparisons with fusion-positive ARMS subsets. Additional cases of “true fusion-negative” ARMS must be identified using the assays described in this paper to permit a detailed examination of the clinical phenotype and corresponding genetic features of this novel subset.