Sarcoma with true epithelial differentiation secondary to irradiated glioblastoma

Glioblastoma multiforme rarely shows true, immunohistochemically confirmed, epithelial differentiation. Furthermore, radiotherapy may induce cerebral sarcomatous tumors, and postsurgery glioblastoma irradiation may give rise to secondary gliosarcomas. We report a case of a 48-year-old male operated on a primary glioblastoma, followed by radiotherapy. A local recurrence occurred 23 months later that was operated too, and a second diagnosis of a fibrosarcoma with true epithelial differentiation was made. Primary systemic neoplasms were largely excluded. The patient died shortly after, and postmortem showed another cerebral dural-attached mass corresponding to a sarcoma without epithelial differentiation, and leptomeningeal seeding composed of malignant epithelial elements only. Cytogenetics, however, disclosed the second tumor to be similar to the primary one.


Introduction
Glioblastoma multiforme with true epi thelial differentiation (TEDGBM), including the presence of squamous cells with epithelial whorls, is one of its rarer subtypes, usually requiring immunohistochemical evidence of both components for diagnosis [1]. However, this differentiation may also be seen in glio sarcomas (GS) [2] and this raises the need for considering the differential diagnosis with metastases of carcinoma and the rare "colli sion" tumors, in which two histologically dis tinct tumors temporally and topographically coexist in the nervous system [2,3,4].
The causal relationship between radio therapy and nervous system tumors is well known [5,6], and criteria for radiationin duced tumors are established [7]. Secondary GS may occur either after a radiotherapy treated GBM [6,8], or after cranial radia tion therapy without a previous history of malignant glioma. Moreover, postcranial irradiationrelated pure sarcomatous tumors have also been described [8].
We aim to describe a case of a recurrent tumor in a patient with a GBM submitted to radiotherapy, in which the complete tumor parenchyma had been replaced by an appar ent fibrosarcoma (FS) displaying extensive squamous cell differentiation tissue, without postmortem evidence of another systemic tu mor. We also try to elaborate on the possible subjacent etiopathogenesis.

Case history
A 48yearold male presented with head aches and behavioral changes; the brain MRI (images not available) showed a contrast enhanced intraaxial mass in the left posterior temporal lobe. A craniotomy was performed with total macroscopic removal of the lesion. After neuropathological diagnosis, the adju vant treatment consisted of radiotherapy (60 Gy) and chemotherapy with temozolomide. However, 23 months later, a new MRI ( Figure  1) disclosed a local recurrence and the patient underwent a second surgery. Following the second neuropathological examination, other primary systemic neoplasms were excluded. A ventriculoperitoneal shunt was placed 4 months later due to an acute hydrocephalus but the pa tient died of pulmonary thromboembolism.
Molecular cytogenetic studies were performed in both tumor samples by high resolution chromosomal comparative ge nomic hybridization (HRCGH) analysis according to a previously described protocol [28]. FISH analyses on tumor imprints were performed with Vysis (Abbott Molecular, Des Plaines, IL, USA) locus specific (LSI) probes: 1p36/1q25; EGFR/centromere of chromosome 7; PTEN/centromere of chro mosome 10 and 19q13/19p13 according to the manufacturer's recommendations.
The tumor from the first surgery ( Figure  2) was composed of neoplastic, poorly dif ferentiated, pleomorphic, GFAP immunore active astrocytes, with nuclear atypia, high mitotic activity, proliferative index higher than 10%, abundant vascular endothelial proliferation and extensive coagulative ne crosis. Epithelial markers were all negative. The diagnosis of GBM was made.
The tumor from the second surgery (Fig  ure 3) showed spindleshaped, bundledis posed, rich in reticulin, vimentine immuno reactive, malignant elements with moderate mitotic activity, proliferative index higher than 10% and extensive coagulative necrosis. In small islands, sharply separated from the sarcomatous tissue, nests of epithelial cells containing irregularly shaped nuclei with mitotic figures, with squamous differentia tion and pearls of keratin were elicited. This epithelial component disclosed immunoreac tivity for epithelial markers only (AE1 AE3, CK14 and EMA), and the proliferative index was higher than 30%. Finally, nests of very few, GFAP immunoreactive, astrocytic cells, with no mitotic figures or Ki67 nuclei immu noreactivity, could also be elicited scattered  throughout the sarcomatous tissue, although mainly at its periphery. The diagnosis of a FS with TED was made.
Complex and very similar chromosomal imbalances were observed in both tumors and are described in Table 1. FISH results confirmed HRCGH data, namely that none of the tumor samples presented EGFR locus amplification.
General autopsy failed to show any tu mor. Macroscopically, there was a right pa rietal lobe, duralattached mass, besides left temporal parenchymal changes.
Microscopic evaluation (Figure 4) showed leptomeningeal tumor dissemination through out the neuroaxis by epithelial, AE1 AE3 and CK14 immunoreactive elements only. The pa renchymal lesion was composed mainly of co agulative necrosis, with widespread dystrophic microcalcification deposits and reactive astro cytosis, suggesting radiationrelated changes. The parenchyma duralattached mass revealed the same sarcomatouslike tumor as that of the second surgery, but neither epithelial nests nor focal glial cells could be elicited.

Discussion
As mentioned, TED in GBM or GS should be differentiated from epithelioid (epithelial like cells) and adenoid (compactly arranged cells, occasionally with pseudoglandular/ cribiform spaces) GBM, both lacking immu noreactivity for epithelialspecific markers, through immunohistochemical confirma tion of this differentiation [1]. Indeed, in the first report of this epitheliallike malignant astrocytic pattern in GS, the presence of transitions from this pattern to neoplastic as trocytes and to their GFAP immunoreactiv ity is mentioned [9]. A few cases similar to the ones of Kepes et al. were subsequently reported in GBM under different names, all having in common epithelioid, GFAP immu noreactive arrangements of neoplastic astro cytes [1,10,11,12,13,14,15], and several mechanisms were proposed to these distinct tumor cell patterns in GBM and GS [10].
TED was first reported by Mørk et al. [2], both in GBM and GS, but this seems to be a very rare occurrence [1,16,17], prompting the differential diagnosis with a metastatic carcinoma [9] and the socalled "collision" tumor [2,3,4], and with several primary brain tumors [1]. Despite the fact that cytogenetics in our case showed the two apparently differ ent tumors to be the same, in what the asso ciation of sarcoma and TED is concerned, the epithelial sarcoma is a well known systemic neoplasia. Two cases of this tumor with cere bral metastases have been described [18,19] but, from the clinical point of view, they were, as expected, very different from our case.
Conventional and molecular cytogenetic studies have demonstrated that GBM and GS have a very specific chromosomal profile which is characterized by concurrent trisomy 7 and monosomy 10 with frequent additional gain of chromosomes 1q, 19 and 20, and losses of chromosomes 9 and 22 [20,21,22]. Molecular genetic evaluation on TEDGBM or GS [1] has also shown a subset of molecu lar patterns of GBM with various degrees of epithelial morphology.
In our case, HRCGH and FISH studies revealed that the two tumors had very similar cytogenetic alterations, namely, both cases presented +7, +19, +20 and loss of 10. These findings thus point out that the patient's sec ond neoplasm should be identified as GBM derived, and exclude the possibility of clas sifying it as a metastatic carcinoma or a pure FS. The cytogenetic analysis of metastatic carcinomas in the brain has shown that these tumors share the same karyotype alterations of their primary tumor counterparts [23]. FS are a poorly studied group of tumors and, to the best of our knowledge, there are no cytogenetic reports of a primary FS of the brain. In adults, FS analysis revealed that the large majority of these neoplasias presented hypodiploiddiploid chromosome numbers and multiple aberrations, where the only nonrandom chromosome changes described were loss of 9p23pter and 10q23qter [24]. Hence, cytogenetics is of paramount impor tance in situations like the present one, and should be performed routinely [1,13,25].
The present case has the unique feature of the second operated tumor disclosing a com pletely different histological type from the first one. Indeed, the possibility of sampling error may exist, although we were cautious enough  and studied a quite representative tissue speci men; on the other hand, we are convinced that, given the small islands of Ki67 immu nonegative astrocytes spreading throughout the second tumor, this component should be considered entrapped reactive astrocytes and not neoplastic ones. Perry et al. [8] described some cases of GS in which the sarcomatous component grew and dominated the histo pathological feature, but it is not mentioned whether they were primary or secondary GS. Given the subarachnoid seeding in postmor tem examination, and having found no other tumor in any of the previous MRI, we may speculate that the duralattached mass should have occurred after radiation, and, accord ingly, should be considered a metastasis of the second, FSlike tumor. Regarding the absence of the epithelial component in this dural mass, we could argue that metastases can underex press the primitive tumor's full histological pattern. However, we have no explanation for the exclusivity of the epithelial component found in the leptomeningeal seeding. The pathogenesis of the aberrant TED in GBM or GS remains unexplained and is poorly addressed in the literature. A hypo thetic role of the sarcomatous component in inducing an epithelial pattern was admitted by Kepes et al. [9], but as mentioned, their cases did not concern TED. As quoted by Mørk et al. [2], in experimentally human smallcell glioblastoma transplanted into mice, the presence of abundant connective tissue stroma in the adenoid mucinproduc ing foci of that tumor was noticed, suggest ing the possible inductive role of the mes enchymal component in the development of that aberrant form of differentiation.
We have no better etiological explana tion for the histological differences between both tumors but the previous irradiation. The causal relationship between a brain tumor ir radiation and the induction of a neoplasm de novo has been reported for meningioma, cerebral FS and other sarcomatous variants, and, rarely, for GBM [6]. The criteria for radiationinduced tumors are that: it should appear in the area of irradiation; a latency pe riod of years should occur between irradiation and the diagnosis of the tumor; it should have been absent prior to the irradiation; and the tu mor de novo should be of a histological type distinct from the previous one [7]. Han et al. [26] reported 30 cases of secondary GS after GBM, 25 having received both externalbeam irradiation and chemotherapy and 3 radiother apy alone, but histopathogical data are men tioned. The same first author had previously reviewed 12 cases of secondary GS and 12 of radiationinduced GS (without previous di agnosis of GBM), comparing clinical and ra diological presentation, response to treatment and pathogenesis [6]. Patients underwent a mean irradiation dose of 54.8 Gy, and the mean time from irradiation to GS diagnosis was of 44.8 weeks. The irradiation doses and latency periods were similar and shorter, re spectively, for those GS in patients in whom a previous glial component was already present as compared with patients in whom a previous malignant glioma was absent, highlighting the potential role of irradiation in facilitating secondary GS. In our case, the latency period between irradiation and tumor recurrence was more than twice the above mentioned period, suggesting that GS harboring predominantly or exclusively a mesenchymatous compo nent may take longer to develop. Interesting enough and slightly approaching this case to our one, extracranial metastases from a GS in which only the sarcomatous component was detected have been reported [27].
The pathogenesis of both forms of sec ondary GS should be considered similar to the one of primary GS [6]. Moreover, radia tion should potentially facilitate the process of: 1) GBM converting local or circulating mesenchymal cells into sarcoma, 2) sarcoma converting local or circulating stem cells into malignant glial cells, or 3) one stem cell lin eage giving rise to both malignant glial and mesenchymal elements. Furthermore, radia tion could induce a simultaneous genesis of glioma and sarcoma from the same progeni tor stem cell, giving rise to a radiationin duced GS [6]. We may speculate that any im balance in one or more of these mechanisms in the sense of a mesenchymal predominance or exclusivity, potentially genetic in cause, could give rise to a secondary radiationin duced FSlike tumor, instead of a classic GS.

Conclusion
Postsurgery radiotherapy in GBM may increase tumor recurrences histomorphologi cally very distinct from the primary one, and cytogenetics are of upmost importance in or der to diagnose similar or different tumors. The etiopathogenesis of this histomorpho logical discrepancy is a matter of discussion. When a true epithelial component is added de novo, this fact should prompt more frequent ly the differential diagnosis with metastases of carcinoma or a collision tumor. FSlike tu mor with TED should also be considered in the list of these secondary neoplasias.