The present study was designed to address a number of questions about OFMT, the first (and perhaps most important) of which is: “Do malignant OFMT exist, and if so, can we reliably recognize them?” We believe our study to answer both of these questions in the affirmative. In our opinion, histologically malignant OFMT clearly exist. Such tumors are defined by the presence of clear-cut areas of morphologically typical OFMT juxtaposed to areas showing high nuclear grade, high cellularity and elevated mitotic activity, while maintaining the overall cytoarchitectural features of OFMT (e.g., lobularity, fibromyxoid matrix, uniform cell-cell spacing and non-pleomorphic round to ovoid cells. Indeed, we have yet to see a completely convincing example of histologically malignant OFMT in which areas of ordinary OFMT were not present, either in the primary tumor or (even more convincingly) in an earlier presentation.
The relationship between typical and malignant OFMT is also supported by our gene expression profiling and immunohistochemical data. We observed very similar expression profiling patterns in both typical and malignant OFMT when compared to nerve sheath myxoma/schwannoma (e.g., upregulation of EAAT4
and downregulation of PMP22
), as well as no significant differences when typical and malignant OFMT were directly compared. Immunohistochemically, typical and malignant OFMT showed roughly similar frequency of expression of the various markers tested, albeit with diminished expression of S100 protein in malignant OFMT, and an increased frequency of focal, weak expression of epithelial markers (e.g. cytokeratins and EMA). Diminished S100 protein expression in malignant OFMT likely reflects malignant progression in the neoplastic cells, akin to the diminished CD34 expression that often accompanies fibrosarcomatous transformation in dermatofibrosarcoma protuberans (1
). In particular, we believe that our finding of a “mosaic” pattern of INI-1 protein loss in identical percentages of typical and malignant OFMT (discussed further, below) argues persuasively in favor of their relationship, as this is a highly unusual pattern of INI-1 protein expression, previously reported only in rare schwannomas associated with familial schwannomatosis (26
) and in some synovial sarcomas (19
). Obviously, different results would be expected if malignant OFMT represented instead other sarcoma types. Given our results, we believe the prior suggestion by Miettinen and colleagues that histologically and clinically malignant OFMT do not exist reflects their arbitrary definition of OFMT as a tumor without atypical features (21
Our data support the validity of the risk stratification system for OFMT proposed initially by Folpe and Weiss in 2003 (13
), inasmuch as we observed clinically malignant behavior (e.g., aggressive local recurrences and distant metastases) only in tumors fulfilling criteria for malignant OFMT. The relatively high percentage of malignant OFMT (32%) in the present study likely reflects the unique nature of our consultation practice. We did not observe any metastases in cases classified as typical or atypical OFMT, although metastases have been previously documented in 4% and 6% of cases falling into those categories, respectively (13
). Metastases were not observed in typical OFMT studied by Miettinen et al, although local recurrences were present in 22% of patients with follow-up (21
). Clinically malignant behavior in OFMT has also been documented in several other small studies, typically in cases showing histologic features suggestive of malignancy (18
). Putting together the results of the present study with those of earlier studies, it is likely that the metastatic risk of typical OFMT is less than 5%, supporting the current WHO classification of OFMT as an “intermediate (rarely metastasizing)” mesenchymal tumor (12
). Cases fulfilling criteria for malignant OFMT should, however, be regarded as high-grade sarcomas.
Turning to the question of the line of differentiation taken by OFMT, our data suggest that OFMT show a “scrambled” phenotype, with limited expression of schwannian and cartilaginous markers, as previously noted, as well as expression of a variety of neural markers, a novel finding. Although schwannian differentiation has generally been favored in OFMT, based on ultrastructural findings, such as the presence of well-developed, occasionally reduplicated external lamina, and S100 protein expression (8
), our gene expression data does not support this. Both by cluster analysis and principal component analysis OFMT were clearly distinct from nerve sheath myxoma and schwannoma, both highly differentiated schwannian neoplasms. Expression of some Schwann cell-related genes, such as PMP22
, was in fact down-regulated in OFMT as compared with nerve sheath myxoma/schwannoma. Cartilaginous differentiation has also been suggested in OFMT, based on ultrastructural features such as irregular cell borders with short processes and intracellular microfilaments, and S100 protein expression (10
). Although we were not able to compare OFMT to cartilaginous tumors by gene expression profiling, we did identify collagen II production in 2 of 5 cases studied by proteomic methods, suggesting that at least some OFMT may show limited cartilaginous differentiation. S100 protein expression in OFMT, shown in this study both by proteomic study and by immunohistochemistry, might also suggest some element of schwannian or cartilaginous differentiation in these tumors, although a wide variety of non-schwannian/ non-cartilaginous cell types may of course express S100 protein (36
). Finally, our findings of only very infrequent cytokeratin or smooth muscle actin expression would seem to offer little support for the notion of myoepithelial differentiation in OFMT.
Surprisingly, we found OFMT to express some neuron-associated genes and proteins, by gene expression profiling, proteomic study and immunohistochemistry. DNA microarray identified a neuron-associated gene, EAAT 4
, to be expressed at high levels in OFMT as compared with schwannian tumors. EAAT4
is a member of the high affinity glutamic acid and neutral amino acid transporter family, expressed principally in the cerebellar cortex (7
). We were able to confirm EAAT4 protein expression in 80% of OFMT by immunohistochemistry. Our microarray data also identified another neuron-related gene, HuC
, as being over-expressed at a 4-5 fold level, with a highly significant p-value (data not shown). However, we were not able to identify a HuC antibody applicable to FFPE tissues, and were thus unable to validate this finding. We did, however, find limited expression of the well-established neuronal markers neurofilament protein and CD56 (neural cell adhesion molecule) in 75% and 41% of tested OFMT, respectively. Finally, proteomic study identified abundant katanin, a neuron-associated microtubule severing protein (2
) and versican, a neuron-associated proteoglycan (3
) in tested cases of OFMT. Although it is obviously premature to ascribe neuronal differentiation to OFMT, in the absence of any convincing light microscopic or ultrastructural supporting evidence, these findings are intriguing. Conceivably OFMT may be showing limited, aberrant expression of neuronal markers due to yet to be defined, dysregulating, upstream genetic events. The significance of expression of MUC4, a mucin-related protein expressed by various epithelia, in OFMT is obscure.
is a putative tumor suppressor gene located on chromosome 22q11.2 which encodes a protein which is expressed essentially in all nucleated cells. Homogeneous loss of immunoreactivity has been identified in most cases of epithelioid sarcoma,(6
) atypical teratoid rhabdoid tumor, (4
) rhabdoid tumor of the kidney and malignant extrarenal rhabdoid tumor (33
), with mutations of INI-1
thought to play a pathogenetic role in tumorigenesis (23
). As noted above, we observed an unusual loss of INI-1 protein expression in 30-60% of neoplastic cells (“mosaic pattern”), a finding previously reported only in schwannomas associated with familial schwannomatosis and in some synovial sarcomas (19
). Our FISH studies for INI-1
suggest that 22q may carry genes with some possible pathogenic relevance in this tumor. Most examined cases showed aneuploidy for chromosome 22 in a significant population of cells suggesting a possible pathogenic role for INI-1
in OFMT. While none of the 7 cases we evaluated were found to have a homozygous deletion for INI-1
, 5 cases showed a hemizygous deletion of both INI-1
(the control probe) on average in more than 50% of cells, and 3 of the 5 had a second population of cells showing 2 signals for INI-1
and 1 signal for the control, suggesting loss of one copy of the 22q telomeric region. However, the molecular events leading to the loss of expression in a subpopulation of tumor cells are unclear. In the heterozygous state for the gene deletion, we hypothesize that epigenetic events such as post translational modifications, or genetic aberrations that are not detectable by FISH such as small deletions or mutations could contribute to the loss of function of the INI-1
on the intact chromosome. On the other hand, the finding of a second existing population of cells with the loss of distal 22q without INI-1
loss raises the possibility that other genes in this chromosomal region, such as the NF2
gene, implicated in neurofibromatosis type II (15
), may play a crucial role in the pathogenesis of OFMT. In any event, we have been able to confirm aneuploidy for the long arm of chromosome 22 in a subset of tested cases, strongly suggesting some role for the INI-1
or an adjacent locus in the pathogenesis of OFMT. Interestingly, mosaic pattern loss of INI-1 was seen both in morphologically and clinically typical OFMT, as well as in malignant OFMT. No difference in behavior was noted in tumors showing retained INI-1 expression versus those with mosaic pattern INI-1 loss. We are not aware of any reports of OFMT arising in the setting of familial INI-1
gene mutations, nor were such histories present in any of our cases.
In conclusion, we believe the results of our study amply confirm the existence of malignant OFMT, as well as the validity of the current risk stratification system for OFMT. Although the line of differentiation taken by this rare tumor remains enigmatic, expression of neuron-related markers in OFMT appears to be significantly more common than is expression of schwannian or cartilaginous markers. The significance of expression of neuronal markers in OFMT remains to be fully elucidated. EAAT4 may prove to be a novel diagnostic marker of use in the differential diagnosis of OFMT, although study of large numbers of potential OFMT mimics is required before these can be recommended. MUC4 is unlikely to be a useful marker of OFMt, particularly as it is known that MUC4 is frequently expressed in at least one potential mimic of OFMT, low-grade fibromyxoid sarcoma (9
). The role of INI-1 loss and 22q aneuploidy in OFMT also warrants further study.