Diagnosing Alveolar Rhabdomyosarcoma: Morphology Must Be Coupled With Fusion Confirmation

Following its initial description as a distinct histologic variant in 1956 by Riopelle and Thériault,¹ and in 1958 by Horn and Enterline,² Enzinger and Shiraki³ reported on a series of 110 cases of alveolar rhabdomyosarcoma (ARMS) that stands to this day as the classic description of this clinically aggressive entity. Patients were between 10 and 20 years of age with tumors that most commonly arose in the extremities or the perirectal and perianal areas. Tumors grew rapidly before their initial diagnosis and had an equally aggressive course, subsequently, with a median survival of fewer than 9 months. Virtually all patients died of their disease within 4 years of diagnosis from regional nodal and distant metastases to the lungs, pancreas, bones, heart, adrenal glands, and other visceral organs. The hallmark histologic feature of this uniquely aggressive variant was the demonstration of a pseudoglandular or pseudoalveolar architecture, yet even those early reports described challenges in establishing the diagnosis primarily when similar small, apparently undifferentiated round cells were seen without the defining architectural features.

To begin to define the optimum treatment for children and adolescents/young adults younger than 21 years of age with rhabdomyosarcoma (RMS), in 1972, the investigators from the soft tissue sarcoma committees of Cancer and Leukemia Group B, Children's Cancer Study Group, and the Southwest Oncology Group formed the Intergroup Rhabdomyosarcoma Study Group (IRSG) and opened the first IRSG study (IRS-I).⁴ Five-year survival for the 686 patients enrolled on this study from 1972 to 1978 was 55%. Although primary site and extent of initial surgical resection (clinical group) were identified as significant prognostic factors, the inferior outcome for patients with ARMS was not clearly independent of other risk factors. Hidden, however, within the larger results of this trial was the observation that the proportion of patients with alveolar histology and metastatic disease at diagnosis was significantly higher (32%) than for those with other stages of disease (overall, 20% of the total cases and ranging from 12% for patients with unresected [group III] disease to 24% for patients with microscopic residual [group II] disease). Virtually identical findings were seen in 999 patients enrolled on the IRS-II trial conducted from 1978 to 1984⁵; again, 20% of all patients had ARMS with an over-representation of ARMS (33%) among those with distant metastases at diagnosis. As in IRS-I, a site- and clinical group–independent adverse prognostic significance of alveolar histology was not clearly defined, but improved outcome with intensification of therapy for patients with extremity ARMS was seen.⁶

Beginning with the IRS-III trial conducted from 1984 to 1991, alveolar histology was defined as an adverse risk factor for which stratification of therapy was indicated.⁷ As in IRS-I and IRS-II, slightly fewer than 20% of the 1,062 patients had (centrally reviewed and agreed on) alveolar histology; however, 56 (37%) of 150 patients with metastatic disease had ARMS. A significantly improved outcome for patients with clinical groups I and II ARMS treated with more intensive systemic therapy was seen. Stratification for histologic subtype was not a feature of the IRS-IV study conducted from 1991 to 1997⁸; as with the first three IRS studies, a remarkably consistent 20% of all cases of nonmetastatic disease were found to be ARMS on central path review. Histologic subtype was found to be prognostically significant in this study, with an estimated 3-year failure-free survival (FFS) of 83% for patients with embryonal RMS versus 66% for patients with alveolar RMS. As before, however, the prognostic significance of alveolar histology independent of site and other clinical features was uncertain.^9–11 Similar confusion exists about the independent prognostic significance of alveolar histology for patients with metastatic RMS.¹²

Further complicating the matter was the change, and then the change back again, in the criteria by which a diagnosis of ARMS was made over the course of the past two IRS studies from one where the demonstration of any alveolar component was sufficient to call a tumor ARMS (as was the case for the fifth-generation IRS studies, Children's Oncology Group Study D9803 and D9602, where an unprecedented 45% of patients with intermediate-risk tumors treated on D9803 had alveolar RMS on central pathology review¹³), to one where the majority of the tumor has to demonstrate alveolar features, as is now the case in the current sixth-generation IRS studies. Although a defining reciprocal translocation encoding a fusion of PAX3 or PAX7 with the FOXO1 gene (ie, FKHR) is known to exist in the vast majority of cases of alveolar RMS, approximately 20% of histologically alveolar tumors contain no demonstrable translocation, and molecular diagnostic techniques are neither uniformly available nor uniformly applied.^14,15

What, then, is an oncologist to do or say when sitting down with the family of a child with newly diagnosed RMS? Does histologic subtype matter? Current-generation United States Cooperative Group RMS trials (Children's Oncology Group studies ARST 0331 for low-risk RMS and ARST 0531 for intermediate-risk RMS) say it does, but only for some patients, with an upgrading of risk group from low to intermediate, and a corresponding intensification of therapy for children with ARMS, but no change in therapy for patients with unfavorable site ARMS. Again, what is an oncologist to do? How should histologic subtypes of RMS be defined and, more importantly, does histology matter?

In this issue of Journal of Clinical Oncology, Williamson et al¹⁶ describe compelling results that will at least answer the first question: when is an alveolar RMS truly an alveolar RMS? The authors interrogated the fusion gene status of 210 appropriately reviewed and annotated RMS specimens divided into three groups: embryonal RMS (ERMS; n = 77); fusion gene–positive alveolar RMS (ARMSp; n = 94); and fusion gene–negative ARMS (ARMSn; n = 39), the latter comprising a group of tumor samples that appeared histologically to be alveolar, but did not have an identifiable PAX-FOXO1 translocation. The authors demonstrate convincingly that ERMS and ARMSn are one and the same; that is, they arise in the same locations and have a comparable frequency of metastases—distinct from ARMSp. Most importantly, they have indistinguishable outcomes with therapies that were not stratified for histologic subtype. Earlier reports provided evidence that translocation-negative ARMS were genetically similar or indistinguishable from ERMS, although whether they were clinically identical as well was less clear.¹⁷

Using expression profiling data from 101 tumors, Williamson et al¹⁶ confirm the sharp differences in expression profiles previously reported between ARMSp and ERMS^18–20 and use these to definitively show that ARMSn are transcriptomally indistinguishable from ERMS. Turning to genomic structural alterations, the differences are equally sharp. Importantly, the article by Williamson et al¹⁶ provides the largest high-resolution genomic copy number data set available to date for pediatric RMS, based on genome-wide copy number profiling data on 128 samples. Distinctions that were previously based on lower resolution studies^21,22 with small sample numbers are now firmly established. Most striking is gain of chromosome 8, seen in 55% to 74% of ARMSn and ERMS but in no cases of ARMSp (0/50). As such, this could argue for adding a simple fluorescence in situ hybridization (FISH) assay for chromosome 8 to the molecular diagnostic work-up in this clinical setting. The biology of these whole-chromosome gains of 8 remains largely unclear, despite FGFR1 being more selectively amplified in rare cases.¹⁶ Conversely, certain gains are typical of ARMSp, but not of ARMSn or ERMS, such as amplification of the fusion gene itself (more often observed for PAX7-FOXO1) and gains of MYCN and CDK4, all previously noted in smaller studies.^21,23,24 Taken together, the clinical and genomic data make a compelling case for the lack of relationship between ARMSp and ARMSn. In other words, making the diagnosis of ARMS in the absence of molecular confirmation of a PAX-FOXO1 fusion is no longer acceptable.

How to incorporate the findings of this article is clear. All tumors that have histologic evidence of alveolar features (cytologically and/or architecturally) should be molecularly evaluated. Tumors that appear alveolar under the microscope, but do not have an identifiable PAX-FOXO1 translocation, should be classified as embryonal. FISH is perhaps the most robust molecular diagnostic technique that can be applied to most tumor samples, even those procured under less than optimal conditions, for detection of a FOXO1 translocation. While this technique can confirm the presence or absence of a PAX-FOXO1 translocation, it does not provide information about the partner gene, PAX3 or PAX7, involved in the translocation—another piece of information that may define distinctive clinical behavior and prognosis^25,26 and that could be useful in future stratification efforts. Thus, it may be desirable to couple FOXO1 FISH with reverse transcriptase polymerase chain reaction for PAX3-FOXO1 and PAX7-FOXO1, with the former used in parallel or as a backup if reverse transcriptase polymerase chain reaction is technically unsatisfactory. Fortunately for this new standard, ARMSp that contain variant fusions other than PAX3-FOXO1 or PAX7-FOXO1 appear exceedingly rare.^18,27

For now, for families of children with newly diagnosed RMS and the doctors who treat them, the results of this important article will go a long way toward reducing the risk that children with fusion-negative low-risk ARMS will receive inappropriately overly intensive therapy. These results will also permit a cleaner interpretation of the prognostic significance of alveolar histology among intermediate-risk patients treated with identical systemic therapy by weeding out the approximately 20% of patients with ARMSn whose outcomes are more favorable than those with ARMSp. Just longer than 5 decades after its initial description, the diagnosis of ARMS has fully entered the molecular era. One can only hope that additional molecular determinants of outcome and identification of clinically exploitable therapeutic targets will follow soon.