Transdifferentiation of neoplastic cells

Medical Hypotheses (2001) 57(5), 655±666 & 2001 Harcourt Publishers Ltd doi: 10.1054/mehy.2001.1435, available online at http://www.idealibrary.com on

1

Transdifferentiation of neoplastic cells Z. Zhang,1 X.-M. Yuan,2 L.-H. Li,1 F.-P. Xie1 1

Department of Pathology, Dalian Medical University, Dalian, China; 2Division of Pathology II, LinkoÈping University, LinkoÈping, Sweden

Summary Transdifferentiation is a process in which a stable cell's phenotype changes to that of a distinctly different cell type. It occurs during certain physiological processes and leads to transition of tumor cell phenotypes. The latter process includes neoplastic epithelial±epithelial transition, neoplastic epithelial±mesenchymal transition, neoplastic mesenchymal±epithelial transition and transition between non-neural and neural neoplastic cell. This phonomenon is exemplified in some origin-debated tumors, such as carcinosarcoma, pleomorphic adenoma, synovial sarcoma, Ewing's/ pPNET, and malignant fibrohistiocytoma. We propose that differentiation disturbance of cancer cells should include not only undifferentiation and dedifferentiation, but also transdifferentiation as well. Tumor cell transdifferentiation may be influenced or determined by cellular genetic instabilities, proliferation and apoptosis, as well as by extracellular matrix and growth factors. & 2001 Harcourt Publishers Ltd

INTRODUCTION

Epithelial metaplasia

Transdifferentiation is a process in which a stable cell's phenotype changes to that of a distinctly different cell type. Transdifferentiation may be a kind of differentiation disturbance occurring in neoplasia, causing the tumor cells to express a phenotype different from that of their progenitor (1).

In pathological conditions, mature epithelial cells of one phenotype change into another, a process called epithelial metaplasia. Since the resulting cancer cells are immature, `metaplasia' seems to be a less than accurate term to describe the transdifferentiation of neoplasia, although metaplasia is essentially a transdifferentiation process occurring in non-neoplastic lesions.

NEOPLASTIC EPITHELIAL±EPITHELIAL TRANSITION Normal epithelial transdifferentiation Transition between different types of epithelium may be a normal phenomenon in the adult body. For example, squamous epithelium lining the mouse vagina changes cyclically in response to sex hormones secreted during the normal estrous cycle (2). Received 12 January 2001 Accepted 25 May 2001 Correspondence to: Professor Zhong Zhang, Pathological Center, Dalian Medical University, 465 Zhongshan Road, Dalian 116027, P R China. Fax: ‡86 0411 472 0610

Transition of carcinoma cell types Conversion of carcinoma cell type, resulting in heterogenicity of the tumor, is not uncommon. In some cases, dual-differentiation was found even in individual cells. The amphicrine carcinomatous cell may exhibit neuroendocrine features at its base, and mucus production and secretory activity in the apical portion (3). Glandular, squamous, and neuroendocrinal phenotypic features have been found in the same carcinomatous cell (4). Chemically induced proliferation of Clara cells in bronchioli of the hamster or mouse may initially form tumors composed of Clara cells, which subsequently are converted into adeno- or squamous carcinoma (5,6). Experimental proliferation of hamster bronchial

655

656 Zhang et al.

neuroendocrine cells may show squamous or excrine differentiation, as well as neuroendocrine features (7). Factors related to the transition of carcinoma cell types In some types of carcinoma (8±10) the poorer the differentiation of tumor cells, the more multidirectional differentiation occurs. This means that conversion of the morphophenotype of carcinoma cells is related to dedifferentiation resulting from tumor progression with subtle sequence changes. Alternatively, changes in the extracellular environment may lead to neoplastic epithelial±epithelial transition. For example, human cholangioadenoma cells transplanted into nude mice convert to squamous carcinoma entirely at the 11th generation (11). When only a single cell of clonalized rat intestinal adenocarcinoma was transplanted into a nude rat or a single cell of clonalized human rectal adenocarcinoma was transplanted into a nude mouse, they both grew to form tumors consisting of columnar, mucoid and neuro-endocrinal cells (12,13). Growth of cecal adenocarcinoma-derived cells on surfaces coated with a combination of type IV collagen and heparan sulfate proteoglycan resulted in both adhesion and induction of endocrine characteristics (14). Endocrine differentiation was found to be highly stimulated when cells were grown on intact, naturally occurring stroma (amniotic membranes, stripped colonic mucosa, fibroblast cells). The cell type of metastatic pulmonary neuroendocrine carcinoma may be squamous and/or glandular (7). Metastasis implies that the neoplastic cells progress to a more dedifferentiated stage, and enter a novel extracellular microenvironment. Both the genetic instabilities and environmental changes lead to neoplastic epithelial± epithelial transition.

NEOPLASTIC EPITHELIAL±MESENCHYMAL TRANSITION Epithelial±mesenchymal transition in the embryo At the blastula stage of an embryo, only singlecell type, proto-epithelium, exists. The transition of proto-epithelium to primitive mesenchyme is the first and most important step in ontogenesis (15,16). Some fetal epithelia still retain the potential to convert into mesenchymal cells. During development of the fetal palate, two palatal shelves fuse at the midline. Epithelial cells on each side of the fused area are converted into fibroblasts. It is suggested that epithelial midline seam provided the trigger for epithelial±mesenchymal transition (15,16).

Medical Hypotheses (2001) 57(5), 655±666

No epithelial-specific transcription factor has been identified to date. Epithelial gene expression may involve only the interplay of ubiquitous transcription factors. Frisch has suggested that when epithelium converts into mesenchyme, tissue-specific transcription factors both activate appropriate mesenchyme-specific gene expression and repress epithelial-type gene expression (17). In Drosophila, genetic studies have identified two transcription factors and their target genes involved in epithelial± mesenchymal transition in gastrulation. One target gene of the transcription factor encodes proteins associated with the regulation of the actin cytoskeleton (18).

Dedifferentiation of epithelial phenotype Co-expression of cytokeratin and vimentin in carcinoma cells should be defined as a change from epithelial to mesenchymal phenotype. Comparing vimentin expression in renal cortical tubules of untreated rats, rats exposed to toxins and carcinogens, and in humans of various ages with or without renal epithelial tumors, showed that dedifferentiation and changes in phenotype of tubule cells were associated with vimentin expression (19). Gould et al. reported that co-expression of cytokeratin and vimentin in mammary carcinoma is more frequent than in normal, hyperplastic lesions and fibroadenoma of breast (20). Domagala et al. found that vimentin was expressed preferentially in tumor cells of high-grade ductal breast carcinomas and in estrogen receptor-(ER) negative and low-ER breast carcinomas with high proliferative activity (21). Eguchi and Okada (22) demonstrated that cloned retinal pigmented epithelial cells from chicken embryo, even at a late stage of development, translate to lens cells. Cells undergoing transdifferentiation must pass through a dedifferentiated state. Such dedifferentiated cells showed no transcriptions of genes that are specific to either pigment cells or lens cells and had an increased expression of c-myc and high mitotic activity (22,23). In numerous instances, dedifferentiation is the essential prerequisite for epithelial±mesenchymal transition, and redifferentiation from the dedifferentiated state is transcriptionally regulated. However, examples of direct transdifferentiation without an intermediated state are also reported (24). Effects of extracellular matrix and growth factors Both extracellular matrix molecules and growth factors play an important role in the process of epithelial± mesenchymal transition. Avian notochord epithelial cells, avian limb ectoderm and embryonic corneal epithelium, adult bovine corneal

& 2001 Harcourt Publishers Ltd

Transdifferentiation of neoplastic cells

endothelium and adult bovine or rodent thyroid follicles can be induced to form mesenchyme when suspended in 3D collagen gels (25). In this process, epithelial phenotypes such as cytokeratin, cell junction and apicalbased polarity, are lost in the elongated and vimentinpositive cells, i.e. fibroblasts, which emigrate from the former surface of the explant. When the cultured cell line (NBT-II) of a rat bladder carcinoma was attached to the collagen type I in the medium, or a substitute for serum, ultroser G, was added to the medium, epithelial-type cancer cells were converted into the fibroblast type. If the inducing factors are removed, the epithelial phenotype may be restored (26). The effects of differentiation, dedifferentiation and redifferentiation require a coordinated network that simultaneously controls cell growth and differentiation. Differentiation-induced cells progress through the G0-arrest cycle, whereby a certain population of cells retains the capacity to de- and redifferentiate and re-enter the cell cycle. In contrast, the rest of the differentiated population enters the irreversible G0 phase (terminal commitment) that finally results in programmed cell death (27). Receptor-mediated actions and via intracellular second messengers cause growth factors to stimulate the cells in the quiescent G0 phase of the cell cycle. When the transcription factors are activated, DNA synthesis is initiated and followed by cell division and dedifferentiation, or further redifferentiation. On the other hand, expression of growth factor receptor may influence the regulation of E-cadherin-mediated cell adhesion. There is increasing experimental evidence to suggest that the epidermal growth factor receptor of tyrosine phosphorylation may lead to inactivation of the E-cadherin±catenin complex in cancer cells, through its interaction with beta- or gama-catenin in the cytoskeleton (28). Hepatocyte growth factor-scatter factor (HGF-SF) has been shown to produce effects similar to those of epidermal growth factor (EGF) with phosphorylation of catenins (29). E-cadherin plays a critical role in the establishment and maintenance of epithelial morphology and differentiation, and loss of E-cadherin may result in epithelial±mesenchymal transition. Cultured neonatel rat hepatocytes (30), mouse mammary gland epithelial cells (NMuMG) (31), nasophanryngeal carcinoma cells (CG-I) (32), and rat bladder carcinoma NBT-II cells (33) undergo epithelial±mesenchymal transition when the medium contains growth factor, such as EGF, TGF-alpha, -beta, and some subtypes of fibroblast growth factor (FGF). Within a few hours of the addition of nanogram quantities of these factors, the peripheral cells of NBT-II epithelial islands start to detach and migrate. Within 10±15 hours, all cells convert into fibroblast-like cells. This change is reversible by withdrawal of the growth factor. Although growth factors and their receptors & 2001 Harcourt Publishers Ltd

657

principally promote cell proliferation and dedifferentiation, some growth factors may also negatively regulate tumor progression (18). Transition of carcinoma to carcinosarcoma Several hypotheses have been proposed to explain the origin of biphasic appearance of carcinosarcoma. Recent findings from comparative molecular analysis of the neoplastic epithelial and mesenchymal components lend strong support to the monoclonal multidirectional histogenesis of these tumors. Using chromosomal inactivation assays, k-ras mutation analysis, p53 mutation analysis and loss of heterozygosity (LOH) studies (34±38), it has been shown that carcinomatous and sarcomatous elements in carcinosarcomas share common genetic alterations, strongly supporting to the monoclonal multidirectional histogenesis of these tumors. It is generally thought that there is a single totipotential stem cell, which gives rise to the multiphasic appearance of carcinosarcoma. However, it is highly doubtful whether a specific totipotential stem cell exists between adult epithelial cells. Usually both carcinosarcomas and carcinomas occur at similar sites. If carcinosarcoma originates from totipotential stem cells, the latter should coexist with the tissue± determined specific stem cell which gives rise to carcinoma. Since it is likely that both types of stem cells at the same location are exposed to the same carcinogen, why is there such disparity in incidence between carcinoma and carcinosarcoma? We know that the primitive mesenchyme derives from proto-epithelium in the embryo, and that normal and neoplastic epithelial cells may convert into mesenchyme. In fact, clinically, carcinoma may progress into carcinosarcoma. Some examples of pure endometrial carcinomacarcinosarcomatous metastases conversion have been documented (39). It also has been reported that carcinosarcoma may arise in duct papilloma of breast or sweat gland adenoma (40,41). In two of 16 cases of gynecological carcinosarcomas with homogeneous or loss of heterogeneous (LOH) patterns, Fojii et al. found the specific patterns of genetic progression to be consistent with sarcomatous components of the neoplasms arising from carcinomatous components (38). The evidence includes additional LOH consistent with genetic progression and diversion from original carcinoma to carcinoma/ sarcoma foci, and additional LOH consistent with genetic progression from original carcinoma to carcinosarcoma only in metastatic foci (38). When the cultured rat hepatocytes were transformed by methyl-nitroso-nitrosoguanidine and then transplanted into homogenous rats, they grew into the following tumor types: epidermoid carcinoma, glandular carcinoma, hepatocarcinoma, undifferentiated carcinoma, sarcoma and carcinosarcoma (42). Medical Hypotheses (2001) 57(5), 655±666

658 Zhang et al.

NEOPLASTIC MESENCHYMAL±EPITHELIAL TRANSITION Mesenchymal±epithelial transition in the embryo During organogenesis, certain primitive mesenchyme may regain epithelial phenotype. Metanephric blastema is composed of highly proliferative mesenchyme. When the primitive epithelial ureteric bud grows into the blastema, a reciprocal inductive interaction occurs between mesenchyme and epithelium, and the primitive renal vesicle (epithelial structure) is formed from the mesenchyme. The formation of renal vesicle reflects de-repression of epithelial promoters, resulting from inhibition of the expression or transcriptional activities of mesenchymal transcription factors. As mesenchyme undergoes transition to epithelium, the intermediate filament is changed from vimentin to cytokeratin, and the non-polarized, loosely associated cells compact together to form polarized, closely associated cells. The compaction of cells is mediated by E-cadherin in a process that is mechanistically analogous to integrinmediated spreading of fibroblasts on an extracellular matrix. E-cadherin expression is crucial for mesenchymal± epithelial transition. The mechanism controlling E-cadherin activation probably involves alteration of the catenin cytoplasmic plaque protein, in the actin cytoskeleton, or the interactions between them (43,44). If, as Birchmeir and Behrens (45) pointed out, the endogenous E-cadherin promoter region is inaccessible in nonepithelial cells, suggesting that the transcriptionally inactivated gene is packaged into chromatin or is otherwise modified, e.g. by methylation, then regulation of epithelial-specific expression of E-cadherin largely is due to specific suppression of promoter activity in nonepithelial cells, rather than specific activation in epithelial cells. Expression of epithelial marker in normal and neoplastic mesenchyme Decidual stromal cells may express CK8, 18 and 19 when cultured in either Condimed or Chang conditioned medium. Similar cytokeratin positivity could also be seen in cultured fetal fibroblasts from skin, chorionic villi and lung, but not in young or adult skin fibroblast cultures using the same cultured conditions (46). Neoplastic cells co-expressing Vim and CK (and/or EMA), or expressing E-cadherin, or with epithelial-like morphology may be found in some primary mesenchymal tumors. Scattered cells which were immunocytochemically positive to CK8 and 18 were found in cultures of Vimpositive, transformed mesenchymal cell lines, including SV40-transformed human fibroblasts, rhabdomyosar coma, rat smooth muscle-derived cells, murine sarcoma Medical Hypotheses (2001) 57(5), 655±666

and hamster BHK-2 cells. When cultures of human SV80 fibroblasts were treated with 5-aza-cytidine, the frequency of CK-positive cells increased significantly, but no cells became positive upon addition of desmosomal proteins (47). Using the monoclonal antibody HECD-1, the extracellular domain of E-cadherin immunoreactivity was identified in a few soft-tissue tumors, especially those with epithelioid features, including pleomorphic rhabdomyosarcomas, clear cell sarcomas, epithelioid sarcomas, synovial sarcomas and diffuse mesotheliomas (48). Treatment with 5-aza-cytidine induces the formation of a low proportion of CK8-containing intermediate filaments in murine fibroblastoidal or myoblastoidal cells derived from teratocarcinoma cells, with co-expression of CK18 and/or 19 in some of those cells. Rare stable cell lines of these cells display true epithelial morphology with typical desmosomes and CKs other than 8 and 18 (49). Effects of tumor suppressor gene and oncogene The Ela gene of adenovirus may act as a tumor suppressor gene in certain human tumor cell lines, not by killing tumor cells, but by converting them into a non-transformed phenotype. Frisch demonstrated that Ela expression partially convert several human tumor cells (rhabdomyosarcoma, fibrosarcoma, osteosarcoma) and fibroblasts into an epithelial phenotype (17). Co-expression of the human Met receptor (met is an oncogene) and its ligand, HGF/SF, in NIH3T3 fibroblasts cause the cells to become tumorigenic in nude mice. Tsarfary et al. reported that the resultant tumors display lumen-like morphology, contain carcinoma-like focal areas with intercellular junctions resembling desmosomes, and co-express CK and Vim (50). Whether the oncogene and tumor suppression gene take part in the mesenchymal±epithelial transition merits further exploration.

HISTOGENESIS OF SOME BI-DIFFERENTIATION TUMORS Pleomorphic adenoma Currently it is believed that the pleomorphic adenomas of the salivary gland originate from intercalated duct epithelium, while myxomatous and chondroid substances are produced from myoepithelium. Recently, Aigner et al. (43) found there were some areas with unequivocal epithelial and mesenchymal differentiation in this tumor tissue. Many areas displayed transitional phenotype cells and the neoplastic epithelial cells may transdifferentiate into mesenchymal cells. & 2001 Harcourt Publishers Ltd

Transdifferentiation of neoplastic cells

Synovial sarcoma±real carcinosarcoma Synovial sarcoma rarely originates from the synovial membrane lining joints and tendon sheaths. The synovial intima is composed of macrophage-like cells and fibroblast-like cells. The flattened, cuboidal or columnar lining cells of spaces in synovial sarcoma tissue show epithelial differentiation to an adenomatous, or, in rare instances, focal squamous, but not synovial intima type. Thus, synovial sarcoma should be identified as a type of carcinosarcoma or carcinoma of mesenchymal tissue (51). Spindle cell type monophasic, biphasic, and glandular type monophasic synovial sarcoma may be regarded respectively as sarcoma, carcinosarcoma and carcinoma in the differentiation spectrum of this neoplasm. Epithelioid sarcoma In view of the immunohistochemical evidence, epithelioid sarcoma may be hypothesized to be a type of mesenchymal carcinoma with simple epithelial differentiation. Electron microscopic photos of the tumor showed a spectrum of cellular differentiation from fibrohistiocytic cells to epithelial-type cells with junctions, microvilli and tonofilaments. Spindle cells showed myofibroblastic and fibroblastic differentiation (52). All the tumors display both vimentin and epithelial markers, just like some poorly differentiated carcinoma cells. In some cases the tumor cells may co-express neurofilament protein, neuron-specific enolase (NSE), or even synaptophysin (53). Karyotypes of the RM-HS1 cell line were reported to contain extensive numerical and structural rearrangements, with up to 24 marker chromosomes (54). Quezado reported allellic loss on chromosome 22q in six of 10 epithelioid sarcomas (55).

659

cells. No evidence of proliferation by the original surface mesothelial cells in response to the asbestos was found (56). Most malignant mesotheliomas are characterized by multiple, complex and heterogeneous chromosomal aberrations. Sreekantaiah et al. reported that the 3p abnormalities, perhaps causally related to the development of this malignancy, and del(6q) might represent a primary change (57). Malignant mesothelioma originate from subserosal mesenchymal cells. As its normal progenitor possesses bidifferentiated potential, this tumor may show a differentiation spectrum from sarcoma to carcinosarcoma and carcinoma. Adamantinoma of long bone Adamantinoma of long bone, composed of epithelial and fibrous components, is closely related to osteofibrous dysplasia (OFD). Most OFDs contain isolated or aggregated keratin-positive cells, which are identical to OFD-like adamantinoma. The recurrent focus of OFD-like tumor with isolated keratin-positive cells may be a classic adamantinoma with abundant epithelium (58). Hazelbag et al. found: (a) focal basement membrane substances in OFD-like areas with keratin-positive cells; (b) increased basement membrane continuity with gain of histological distinction between epithelial and fibrous components; and (c) strong tenascin reactivity directly surrounding well-developed epithelial fields, with weaker staining more distantly (59). Hazelbag et al. suggested that the epithelial component in adamantinoma may be transformed directly from osteofibrous tissue (59). HISTOGENESIS OF EWING'S/pPNET Hypothesis on the origin of Ewing's sarcoma

Malignant mesothelioma The regenerating mesothelium may originate from either the surrounding uninjured mesothelial cell population or from subserosal cells. It has been suggested that subserosal cells are distinct from other connective tissue cells in this ability to co-express Vim and CK, and serve as replicated cells which can differentiate into surface epithelium. To study this phenomenon, both the visceral and parietal peritonea of rats was incised and allowed to heal with resultant proliferation and differentiation of subserosal cells. The proliferated subserosal cells change morphologically from the original spindleshaped, fibroblast-like cells to polygonal surface epithelial cells (56). Following injection of crocidolite asbestos into the peritoneal cavity of rats and mice, the subserosal cells proliferated to form a tumorous nodule, in which the spindle cells differentiated into surface mesothelial & 2001 Harcourt Publishers Ltd

Ewing's/pPNET (Ewing's sarcoma/peripheral primitive neuroectodermal tumor) is a family of bone and softtissue tumors, in which the typical Ewing's sarcoma, with a mesenchymal phenotype, lies at one end of the differentiation spectrum and pPNET, with clear evidence of a neural marker, lies at the other end. Some authors had suggested that Ewing's sarcoma originates from the mesenchymal or primitive form of connective tissue (60). Aurias et al. and Wang-Peng et al. found a coincidental cytogenetic abnormality in ES and pPNET (60). In 1990 several groups of investigators found that, through the addition of agents such as retinoic acid to the cultures, the ES cells may be induced in vitro to express neural markers (60). Currently, many pathologists prefer to support the view that Ewing's/pPNET originates from neural crest cells. The embryonic neural crest has the potential to differentiate into mesenchyme, which could explain the expression of Ewings' mesenchymal phenotype. Medical Hypotheses (2001) 57(5), 655±666

660 Zhang et al.

Against the hypothesis of a neural origin of Ewing's/pPNET Although normal mesenchyme does not express neural differentiation, `neural' markers may be found in mesenchymal tumors. For example, epithelioid sarcoma cells express neurofilament, NSE and synaptophysin protein (61); MFHs have neurofilaments and are neural-associated antigen-positive (62); synovial sarcoma may be immunohistochemically positive to neurofilament 68kd, S-100, Leu-7 and GFAP (63). Molenaar and Muntinghe (64) studied the expression of neural cell adhesion molecules and neurofilament protein isoforms in bone and soft-tissue sarcomas. They found that both neural markers are expressed in rhabdomyosarcoma, leiomyosarcoma, fibrosarcoma, MFH, malignant rhabdoid tumor and fibromatosis, and are N-CAMpositive in synovial sarcoma. Desmoplastic small round cell tumors (DSRCT) may show positive staining with antibodies to NES, Leu-7 and other neural antigens, although they are more likely from mesenchymal tissue such as subserosal cells (65). Thus, the expression of `neural' markers does not prove a neural origin for the tumor. Neuroblastoma originates from the sympatho-adrenal lineage of the neural crest. Its characteristic chromosome changes are deletion of 1p36.2±3, amplification of the proto-oncogene MYCN, and abnormalities of the chromosome number (66). The special cytogenetic aberration found in Ewing's/pPNET is chromosomal translocation resulting in gene fusion. One of the criteria used to diagnose classic neuroblastoma is the presence of Homes± Wright pseudorosettes. However, the number of rosettes seen by most investigators in pPNETs is few or none (60). The differentiation-potential of neuroblastoma shows neural mono-direction. Ewing's/pPNET shows multidifferentiation and may overlap somewhat in appearance with DSRCT, or even alveolar rhabdomyosarcoma, both of which also show characteristic chromosome translocation. From these points, Ewing's/pPNET is more closely related to non-neural derived tumor in phenotype and cytogenetic profile. The discovery of a hybrid tumor comprised of Ewing's/pPNET and DSRCT further demonstrated that DSRCT and pPNET may share a common histogenesis (67). Neural crest cells detach from the dorsal neural tube and migrate, colonize and differentiate into a large variety of derivatives. These include the neurons and glial cells of the peripheral nervous system, endocrine cells, melanocytes and the so-called mesoectoderm, which plays a crucial role in formation of the connective tissues and skeletal structures of the head. Most of the Ewing's/ pPNET were located outside the head, neck and regions unrelated to any specific structures of the peripheral or

Medical Hypotheses (2001) 57(5), 655±666

sympathetic nervous systems. These tumors are thought to originate from abnormally migrated neural crest cells. However, the occurrence of Ewing's/pPNET in adult and aged patients suggests that an implanted neuroectodermal fetal origin of these tumors is unlikely. Fukunaga et al. reported a case of carcinosarcoma with neuroectodermal differentiation which occurred in the uterus of a 54-year-old woman (68). The neoplastic elements included squamous carcinoma, leiomyosarcoma and islands of small- to medium-sized cells with rosette-like formation and immunoreactivity for GFAP, Leu-7, NSE and synaptophysin. Certainly in this case, the cells with neural differentiation did not arise from neural crest cell. Relation between chromosome translocation and morphogenesis in Ewing's/pPNET Ewing's/pPNET is characterized by specific chromosomal translocations, which result in a fusion gene and its encoding chimeric protein. In at least 90% of Ewing's/ pPNET, the translocation is t(11;22) (q24;q12), resulting in fusion of parts of the EWS gene with parts of the FLI-1 gene encoding for transcription factor. In the encoded chimeric protein, the amino-terminal EWS domain is linked to the DNA-binding domain of the transcription factor. In another 5% of cases, t(21;22) (q22;q12) translocation is present, which involves the genes for EWS and ERG. Another rare translocation is t(7;22) (p22;q12), which results in EWS/ETV-1. FLI-1, ERG and ETV-1 all belong to the ETS proto-oncogene family (69). EWS/FLI-1 (or EWS/ERG, ETV-1) is a transcription factor. May et al. demonstrated that EWS/FLI-1 efficiently transformed NIH3T3 cells, but FLI-1 did not (70). Experiments with EWS/FLI-1 deletion mutants indicated that both EWS and FLI-1 domains are necessary for transformation by the t(11;22) translocation product (71). Recently, Teitell et al. reported that NIH3T3 fibroblasts infected with either EWS/FLI-1 or EWS/ETV-1 resembled small round cell tumor microscopically, and showed both epithelial and neuroectodermal phenotypes as represented by CK15 expression, dense junction, and neurosecretory granule formation (72). The altered cells lost both extracellular collagen deposition and RER, indicating a loss of mesenchymal differentiation. Although all the Ewing's/pPNETs tumors harbor an identical model of the chromosomal translocation, its differentiated spectrum includes a mesenchymal phenotype at one end, and a neural at the other. Other small cell tumors of soft tissue, such as DSRCT, polyphenotypic tumors, and even a few cases of alveolar or embryonal rhabdomyosarcoma, may also possess t(11;22) (q24;q12) translocation and the EWS/FLI-1 fusion gene (73,74). From these findings, we deduce that expression of the neural

& 2001 Harcourt Publishers Ltd

Transdifferentiation of neoplastic cells

phenotype in Ewing's/pPNETs is not decided merely by the formation of EWS/FLI-1 or EWS/ERG fusion gene, but also requires the activation of certain genes related to neural differentiation. Difference in origin between cPNET and Ewing's/pPNET There are two categories of primitive neuroectodermal tumors (PNETs): central (cPNET) and peripheral (pPNET) categories (75). A tumor which originates in the brain or spinal cord, such as medulloblastoma, is classified as a central PNET. Adrenal and extra-adrenal neuroblastomas and Ewing's/pPNET (in soft tissue and bones) are examples of a peripheral PNET. Because they occur in these specific locations, central PNETs and neuroblastomas may derive from neural tissue. As for Ewing's/pPNET, we hypothesize that they arise from mesenchymal tissue. In view of the preceding part of this review, the reasons may be summarized as follows: expression of neural markers does not signify a neural origin; the primary location of most tumors is unrelated to any specific structure of the nervous system; its phenotype is distinct from that of neuroblastoma but similar to that of some small cell tumors of the mesenchyme; its characteristic chromosome translocation and fusion gene, with which the infected 3T3 fibroblasts gain neuroectodermal phenotypes and loss of mesenchymal differentiation. It is evident that the function of EWS/FLI-1 or EWS/ERG and EWS/ETV-1 is not only to initiate the transformation of mesenchymal cells, but also to regulate the phenotypic expression of Ewing's/ pPNET. Vasoactive intestinal peptide (VIP) is a neuromodulator which regulates both proliferation and differentiation of neuronal precursors. Fruhwald et al. analyzed cPNET cell lines, cPNET tumors and Ewing's/pPNET, using reverse transcriptase-polymerase chain reaction and Southern hybridization (76). They found that VIP receptor 1 (VIPR1) and VIPR2 are more highly expressed in both primary cPNET tumors and cPNET cell lines. They pointed out that this result may reflect the divergent pathways of cPNET toward neural differentiation and Ewing's/pPNET to mesenchymal cells. The question is how such divergent pathways of differentiation occur. Another possible explanation is that the origin of cPNET is distinct from that of pPNET; the former from neural tissue, and the latter from mesenchymal cells. HISTOGENESIS OF MALIGNANT FIBROHISTIOCYTOMA Malignant fibrohistiocytoma (MFH) is composed of fibroblast-like cells, histiocyte-like cells and an

& 2001 Harcourt Publishers Ltd

661

intermediate cell type (77,78). These neoplastic cells may show myofibroblast-like differentiation, occasionally acquire certain properties of osteoblasts or chondroblasts, and immunohistochemically express CK, epithelial membrane antigen (EMA) neurofilaments (NF) and neural-associated antigens (79±82).

Heterogeneity of MFH Histocyte origin Since the cases studied did not react with anti-monocytemacrophage antibodies, some investigators are moving away from histiocyte-origin hypothesis (83±88). However, other investigators have reported that MFH does express antigenic markers specific for monocytes-macrophages (89±91). Yumoto and Morimoto inoculated SV40-transformed bone marrow macrophages in syngeneic mice to produce a transplantable tumor (92). The tumor was composed of spindle cells with histiocytic function such as storiform arrangement. Some tumor cells present a transitional form between histiocyte-like and fibroblastlike cells. This experiment demonstrated that MFH may be derived through transdifferentiation of histiocytes. Fibroblast origin This theory is supported by substantial evidence: MFH cells expressed mesenchymal antigens (84,85,88). Injection of DMBA into the knee-joint cavity of rats may induce MFHs and fibrosarcomas or synovial sarcomas in and around the joint (93,94). After several recurrences, fibrosarcoma may transform into MFH (95). Fibrosarcoma cells (RFS) derived from a cadmium-induced fibrosarcoma of rat, and RFS cells in different subcultures produced MFH and fibrosarcoma in nude mice and baby rats (95). Non-transformed and transformed fibroblast-like mouse embryo cells can be induced to differentiate into macrophages or histiocytes in a medium supplemented with human plasma (96). These findings demonstrated that histiocyte-like cells could result from transdifferentiation of neoplastic fibroblasts. Other origins The dedifferentiated components of dedifferentiated sarcoma (e.g. dedifferentiated liposarcoma, dedifferentiated chondrosarcoma and dedifferentiated leiomyosarcoma) and sarcomatous cells of carcinosarcoma, e.g. sarcomatous carcinoma of lung, kidney and prostate and carcinosarcoma of breasts, usually look like MFH or fibrosarcoma (97±105). In addition, the individual case of malignant melanoma, gliosarcoma and malignant lymphoma, has the same appearance as MFH on microscopic or even electron-microscopic examination (106±108).

Medical Hypotheses (2001) 57(5), 655±666

662 Zhang et al.

Karyotype progression of MFH The karyotype of MFH is far more complex than that of fibrosarcoma. Cytogenetic evidence of clonal evolution has been found in some tumors (109). More than half the tumors were near triploid or near tetraploid, some were hyperdiploid and even hypodiploid (110). The tumors develop through acquisition of structural aberrations and chromosome loss or through gain of a chromosome (111). Ring, dicentric- and telomeric-associated chromosomes seem to represent early events in the development of MFH (109,112,113). Sometimes, the presence of ring chromosomes is the sole cytogenetic aberration of myxoid MFH (112,113), as well as the sole consistent chromosomal alteration of well-differentiated liposarcoma and dedifferentiated liposarcoma with area of MFH elements (114±116). Ring, dicentric- and telomeric-associated chromosomes result from telomeric erosion and are highly unstable, leading to further breakage and fusion (117). Genberg et al. established two cell lines from two subsequent recurrences of MHF with identical marker chromosome in common. The U-2149 cell line from the second recurrence consisted of mainly fibroblast-like cells. The U2197 cell line from the third recurrence was composed of mainly histiocyte-like cells. Based on their data, Genberg et al. explained the appearance of histiocyte-like cells in MFH as a consequence of chromosomal progression (118).

APOPTOSIS AND TRANSDIFFERENTIATION Apoptosis is a universal cellular suicide pathway in normal elimination and development. Disruption of this normal process resulting in illegitimate cell survival can facilitate cancer development and progression. In histological tumor material apoptotic index (AI) may be used as a measure of the extent of apoptosis. Most often AI is defined as a percentage of apoptotic cells and bodies per total number of tumor cells. A consistent feature in many studies is the positive correlation or association between apoptosis and proliferation. In general, rapidly growing tumors have a greater degree of apoptosis than relatively indolent ones. Some investigators have begun investigating the correlation between apoptosis and neoplastic transdifferentiation. The Bcl-2 family and apoptosis in synovial sarcoma The very recent report of Antonescu et al. (119) indicated that the monophasic types of synovial sarcoma (with SYTSSX2) have a significantly higher mean and median ki-67 labeling index than biphasic types (with SYT-SSX1) (119), and apoptosis as assessed by TUNEL was rarely observed in both types, consistent with prominent expression of the Medical Hypotheses (2001) 57(5), 655±666

anti-apoptosis protein Bcl-2 in almost all cases. Bcl-2 protein expression was strong in spindle cells in contrast to the weak or negative reactivity observed in epithelial cells. Bcl-2-positive spindle cells consistently surrounded the negative epithelial elements, a pattern confirmed by the studies of Antonescu et al. and other investigators (119). The spindle and adenomatous elements in biphasic synovial sarcoma may be analogous to the undifferentiated and well-differentiated elements of carcinoma, confirming that bcl-2 protein expression is indeed different between two areas of the biphasic tumor.

p53 mutation or inactivation, and apoptosis in MFH The p53 tumor-suppressor gene is the most commonly mutated gene in human cancers. Two biological functions are attributed to p53: regulation of cell cycle progression after stress, and regulation of apoptosis. Some evidence indicates that, although p53 mutations are unlikely to be the primary cause, they may exacerbate chromosome instability (120). Pilotti et al. investigated p53 and MDM2 over-expression in 98 lipomatous tumors by immunocytochemistry, as well as by molecular and cytogenetic analysis (121). Among the 74 cases of liposarcomas, 14 cases were of the dedifferentiated subtype in which the differentiated component was well differentiated, while the dedifferentiated area had MFH features in 13 cases. The results show that, in the retroperitoneal well-differentiated±dedifferentiated group, MDM2-mediated inactivation of p53 could be related to the mechanism of transdifferentiation, while, in the nonretroperitoneal WD-DD group, the Tp53 mutations appear to correlate with the dedifferentiation process. Hisaoka et al. studied eight cases of dedifferentiated liposarcoma with myxoid MFH areas (122). The clinicopathologic, cytogenetic and molecular features showed that the differentiated portion of tumors was more closely related to well-differentiated liposarcoma rather than to ordinary myxoid liposarcoma, while the dedifferentiated part was myxoid MFH. The myxoid portions frequently showed a higher proliferative activity, and more MDM-2 and p53-positive tumor cells than in well-differentiated areas. Hisaoka et al. claimed that an altered p53 pathway, including p53 gene mutations and MDM-2-mediated inactivation of p53, may play a role in tumorigenesis of this myxoid subtype of liposarcoma and its progression (dedifferentiation) toward conversion into myxoid MFH (122).

CONCLUSION Differentiation disturbance of cancer cells should include not only undifferentiation and dedifferentiation, but also & 2001 Harcourt Publishers Ltd

Transdifferentiation of neoplastic cells

transdifferentiation. A spectrum of phenotypic features has been identified in various embryonal carcinoma cells, which resemble normal embryonic cells at different stages of development (123). Just like embryonal carcinoma, most cancers are similar in phenotype to their normal forebears, but lower in differentiation, e.g. dedifferentiation or even undifferentiation. On the other hand, in some tumors, as compared to their normal progenitor cells, the morphophenotypes of tumor cells are converted into other types by so-called transdifferentiation. The pathogenesis and progression of neoplasm are closely associated with genetic instabilities, including subtle sequence changes, alteration in chromosome number, chromosome translocation and gene amplification (120). Genetic instabilities are the basis of neoplastic dedifferentiation and transdifferentiation. Neoplastic epithelial± epithelial transition or neoplastic epithelial±mesenchymal transition may be simply the result of subtle sequence changes and alteration in extracellular matrix molecules or stimulation by growth factors, via regulation of cell proliferation, differentiation and programmed cell death. As to neoplastic mesenchymal±epithelial transition in synovial sarcoma and mesenchymal±neural transdifferentiation in Ewing's/pPNET, chromosome translocations (gene rearrangements) are crucial for transdifferentiation. The histological origin of some groups of neoplasms has been debated for decades. Debated tumors include: multidirectional differentiation tumors; tumors with the characteristic phenotype of a certain tissue occurring in an organ foreign to such tissue, e.g. meningioma in lung; and tumors with a phenotype dissimilar to any type of normal tissue, e.g. alveolar soft part sarcoma. The histological appearances of these tumors may result from transdifferentiation of certain neoplastic cells, and their histogenesis may be resolved at the cytogenetical and molecular levels. REFERENCES 1. Zhang Z., Xie F.-P. Transdifferentiation of neoplastic cell. Foreign Medical Science (Oncology) (in Chinese) 1994; 21 (suppl): 8±9. 2. Horvat B., Vrcic H., Damjanov I. Transdifferentiation of murine squamous vaginal epithelium in proestrus is associated with changes in the expression of keratin polypeptides. Exp Cell Res 1992; 199: 234±239. 3. Hammer S. P., Insalaco S. J., Lee R. B. Amphicrine carcinoma of the uterine cervix. Am J Clin Pathol 1992; 97: 516±522. 4. McDowell E. M., Trump B. F. Pulmonary small cell carcinoma showing tripartite differentiation in individual cells. Hum Pathol 1981; 12: 286. 5. Rehm S., Takahashi M., Ward J. M. et al. Immunohistochemical demonstration of Clara cell antigen in lung tumors of bronchiolar origin induced by N-nitrosodiethylamine in Syrian golden hamsters. Am J Pathol 1989; 134: 79±87. 6. Rehm S., Lijingsky W., Singh G. et al. Mouse bronchiolar cell carcinogenesis. Histologic characterization and expression

& 2001 Harcourt Publishers Ltd

663

of Clara cell antigen in lesions induced by N-nitrosobin(2-chloroethyl) ureas. Am J Pathol 1991; 139: 413±422. 7. Gould V. E. Histogenesis and differentiation. A re-evaluation of these concepts as criteria for the classification of tumors. Hum Pathol 1986; 17: 212±215. 8. Xie Z. X., Liao J. H., Zheng P. E. Immunohistochemical and electromicroscopic observation for endocrine and Paneth's cell of gastric carcinoma. Chin J Pathol (in Chinese) 1992; 21: 221±223. 9. Wu W. Y., Sun Y. N., Liu Q. et al. Immunohistochemical study on bladder transitional cell carcinoma containing HCG. Chin J Pathol (in Chinese), 1989; 18: 34±36. 10. Jiang S. X., Yang G. H., Tan Y. S. et al. Study on cellular heterogeneity of bronchiogenic carcinoma. Chin J Pathol (in Chinese) 1991; 20: 178±180. 11. Iemura A., Yano H., Mizogushi A. et al. A cholangiocellular carcinoma nude mouse strain showing histologic alteration from adenocarcinoma to squamous cell carcinoma. Cancer 1992; 70: 415±422. 12. Cox W. F., Pierce G. B. The endodermal origin of the endocrine cells of an adenocarcinoma of the colon of the rat. Cancer 1982; 50: 1530±1538. 13. Kirkland S. C. Clonal origin of columnar, mucous, and endocrine cell lineages in human colorectal epithelium. Cancer 1988; 61: 1359±1363. 14. de Bruine A. P., Dinjens W. N. M., van den Linden E. P. M. et al. Extracellular matrix components induce endocrine differentiation in vitro in NCL-H716 cells. Am J Pathol 1993; 142: 773±782. 15. Fitchet J. E., Hay E. D. Medial edge epithelium transforms to mesenchyme after embryonic palatal shelves fuse. Dev Biol 1989; 131: 455±474. 16. Griffith C. M., Hay E. D. Epithelial-mesenchymal transformation during palatal fusion: Carboxyfluorescein traces cells at light and electron microscopic levels. Development 1992; 116: 1087±1099. 17. Frisch S. M. Ela induces the expression of epithelial characteristics. J Cell Biol 1994; 127: 1085±1096. 18. Thiery J. P., Chopin D. Epithelial cell plasticity in development and tumor progression. Cancer Metast Rev 1999; 18: 31±42. 19. Ward J. M., Stevens J. L., Konishi N. et al. Vimentin metaplasia in renal cortical tubules of preneoplastic, neoplastic, aging and regenerative lesions of rats and humans. Am J Pathol 1992; 141: 955±964. 20. Gould V. E., Koukoulis G. K., Jansson D. S. et al. Coexpression patterns of vimentin and glial filament protein with cytokeratins in the normal, hyperplastic and neoplastic breast. Am J Pathol 1990; 137: 1143±1156. 21. Domagala W., Lasota J., Bartkowiad J. et al. Vimentin is preferentially expressed in human breast carcinomas with low-estrogen receptor and high Ki-67 growth fraction. Am J Pathol 1990; 136: 219±227. 22. Eguchi G., Okada T. S. Differentiation of lens tissue from the progeny of chick retinal pigment cells cultured in vitro: A demonstration of a switch of cell type in clonal cell culture. Proc Natl Acad Sci USA 1973; 70: 1495±1499. 23. Agata K., Kobayashi H., Itoh Y. et al. Genetic characterization of the multipotent dedifferentiated state of pigmented epithelial cells in vitro. Development 1993; 118: 1025±1030. 24. Eguchi G., Kodama R. Transdifferentiation. Curr Opin Cell Biol 1993; 5: 1023±1028. 25. Hay E. Extracellular matrix alters epithelial differentiation. Curr Opin Cell Biol 1993; 5: 1029±1035. 26. Boyer B., Tucker G. C., Valles A. M. et al. Reversible transition towards a fibroblastic phenotype in a rat carcinoma cell line. Int J Cancer 1989; (Supple) 4: 69±75.

Medical Hypotheses (2001) 57(5), 655±666

664 Zhang et al.

27. Hass R. Retrodifferentiation and cell death. Criti Rev Onco-genesis 1994; 5: 359±371. 28. Jawhari A. U., Farthing M. J. G., Pignatelli M. The E-cadherin/ epidermal growth factor receptor interaction ± A hypothesis of reciprocal and reversible control of intercellular adhesion and cell proliferation. J Pathol 1999; 187: 155±157. 29. Crepaldi T., Pollack A. L., Part M. et al. Targeting of the SF/HGF receptor to the basolateral domain of polarized epithelial cells. J Cell Biol 1994; 125: 313±320. 30. Pagan R., Martin I., Alonso A. et al. Vimentin filaments follow the pre-existing cytokeratin network during epithelial mesenchymal transition of cultured neonatal rat hepatocytes. Exp Cell Res 1996; 222: 333±334. 31. Miettnen P. J., Ebner R., Lopez A. R. et al. TGF-beta induced transdifferentiation of mammary epithelial cells to mesenchymal cells: involvement of type I receptors. J Cell Biol 1994; 127: 2021±2036. 32. Chen J. K., Shen L. S., Chao H. H. Correlation of transformation from epithelial to mesenchymal-like morphology and endogenous bFGF levels in human nasopharyngeal carcinoma cells. J Cellular Physiol 1994; 160: 401±408. 33. Valle A. M., Boyer B., Badet J. et al. Acidic fibroblast growth factor is a modulator of epithelial plasticity in a rat bladder carcinoma cell line. Proc Natl Acad Sci USA 1990; 87: 1124±1128. 34. Thompson L., Chang B., Barsky S. H. Monoclonal origins of malignant mixed tumors (carcinosarcomas). Evidence for a divergent histogenesis. Am J Surg Pathol 1996; 20: 277±285. 35. Wada H., Enomoto T., Fujita M. et al. Molecular evidence that most but not all carcinosarcomas of the uterus are combination tumors. Cancer Res 1997; 57: 5379±5385. 36. Kounelis S., Jones M. W., Papadaki H. et al. Carcinosarcomas (malignant mixed mullerian tumors) of the female genital tract: Comparative molecular analysis of epithelial and masenchymal components. Hum Pathol 1998; 29: 82±87. 37. Abeln E. C., Smit V. T. H. B. M. Wessels J. W. et al. Molecular genetic evidence for the conversion hypothesis of the origin of malignant mixed mullerian tumours. J Pathology 1997; 183: 424±431. 38. Fujii H., Yoshida M., Gong Z. X. et al. Frequent genetic heterogeneity in the clonal evolution of gynecological carcinosarcoma and its influence on phenotypic diversity. Cancer Res 2000; 60: 114±120. 39. Sreenan J. J., Hartn W. R. Carcinosarcomas of the female genital tract. A pathologic study of 29 metastatic tumors: Further evidence for the dominent role of the epithelial component and the conversion theory of histogenesis. Am J Surg Pathol 1995; 19: 666±670. 40. Pitt M. A., Swells S., Eyden B. P. Carcinosarcoma arising in a duct papilloma. Histopathol 1995; 26: 81±84. 41. Mckee P. H., Fletcher C. O. M., Stavrino P. et al. Carcinosarcoma arising in eccrine spiradenoma (a clinicopathologic and immunohistochemical study of 2 cases). Am J Dermatopathol 1990; 12: 335±343. 42. Tsao M. S., Grisham J. W. Hepatocarcinomas, cholangiocarcinomas, and hepatoblastomas produced by chemically transformed cultured rat liver epithelial cells. Am J Pathol 1987; 127: 169±181. 43. Aigner T., Neureiter D., Volker U. et al. Epithelial-mesenchymal transdifferentiation and extracellular matrix gene expression in pleomorphic adenomas of the parotid salivary gland. J Pathol 1998; 186: 178±185. 44. Gumbiner B. M. Cell adhesion: The molecular basis of tissue architecture and morphogenesis. Cell 1996; 84: 345±357. 45. Birchmeier W., Behrens J. Cadherin expression in carcinoma: Role in the formation of cell junction and the

Medical Hypotheses (2001) 57(5), 655±666

46.

47.

48.

49.

50.

51. 52.

53.

54.

55. 56.

57.

58.

59.

60. 61.

62.

63.

64.

prevention of invasiveness. Biochem Biophys Acta 1994; 198: 11±26. von Koskull H., Virtanen I. Induction of cytokeratin expression in human mesenchymal cells. J Cell Physio 1987; 133: 321±329. Knapp A. C., Franke W. W. Spontaneous losses of control of cytokeratin gene expression in transformed, non-epithelial human cells occurring at different levels of regulation. Cell 1989; 59: 67±79. Sato H., Hasegawa I., Abe Y. et al. Expression of E-cadherin in bone and soft tissue sarcomas: A possible role in epithelial differentiation. Hum Pathol 1999; 30: 1344±1349. Semat A., Duprey P., Vasseur M. et al. Mesenchymal-epithelial conversions induced by 5-azacytidine: appearance of cytokeratin endo-A messenger RNA. Differentiation 1986; 31: 61±66. Tsarfary I., Rong S., Ressau J. H. et al. The Met proto-oncogene mesenchymal to epithelial cell conversion. Science 1994; 263: 98±101. Chadially F. N. Is synovial sarcoma a carcinosarcoma of connective tissue? Ultrastruct Pathol 1987; 11: 147±151. Fisher C. Epithelioid sarcoma. The spectrum of ultrastructural differentiation in seven immunohistochemically defined cases. Hum Pathol 1988; 19: 265±275. Gerharz C. D., Moll R., Ramp U. et al. Multidirectional differentiation in a newly established human epithelioid sarcoma cell line (GRU-1) with co-expression of vimentin, cytokeratins and neurofilament proteins. Int J Cancer 1990; 45: 143±152. Reeves B. R., Fisher C., Smith S. et al. Ultrastructural, immunocytochemical, and cytogenetic characterization of a human epithelioid sarcoma cell line (RM-HS1). J Natl Cancer Inst 1987; 78: 7±18. Quezado M. M. Alleli loss on chromosome 22q in epithelioid sarcomas. Hum Pathol 1998; 29: 604±608. Bolen J. W., Hammar S. P., McNutt M. A. Reactive and neoplastic serosal tissue. A light-microscopic, ultrastructural, and immunocytochemical study. Am J Surg Pathol 1986; 10: 43±47. Sreekantaiah C., Ladanyi M., Rodriguez E. et al. Chromosomal aberrations in soft tissue tumors. Relevance to diagnosis, classification, and molecular mechanism. Am J Pathol 1994; 144: 1121±1134. Sweet D. E., Vinh T. N., Devaney K. Cortical osteofibrous dysplasia of long bone and its relationship to adamantinoma: A clinicopathologic study of 30 cases. Am J Surg Pathol 1992; 16: 282±290. Hazelbag H. M., Van den Broek L. J. C. M., Fleuren G. J. et al. Distribution of extracellular matrix components in adamantinoma of long bones suggests fibrous-to-epithelial transformation. Hum Pathol 1997; 28: 183±188. Dehner L. P. Primitive neuroectodermal tumor and Ewing's sarcoma. Am J Surg Pathol 1993; 17: 1±13. Gerharz C. D., Moll R., Ramp U. et al. Multidirectional differentiation in a newly established human epithelioid sarcoma cell line GRU1 with coexpression of vimentin cyto-keratin and neurofilament proteins. Int J Cancer 1990; 45: 143±152. Lawson C. W., Fisher C., Gatter K. C. An immunohistochemical study of differentiation in malignant fibrous histiocytoma. Histopathol 1987; 11: 375±383. Noguera R, Navarro S, Cremades A. et al. Translocation 18 in a biphasic synovial sarcoma with morphologic feature of neural differentiation. Diagn Mol Pathol 1998; 7: 16±23. Molenaar W. M., Muntinghe F. L. H. Expression of neural cell adhesion molecules and neurofilament protein isoforms in Ewing's sarcoma of bone and soft tissue sarcomas other than rhabdomyosarcoma. Hum Pathol 1999; 30: 1207±1212.

& 2001 Harcourt Publishers Ltd

Transdifferentiation of neoplastic cells

65. OrdoÂnez N. G. Desmoplastic small round cell tumor II. An ultrastructural and immunohistochemical study with emphasis on new immunohistochemical markers. Am J Surg Pathol 1998; 22: 1314±1327. 66. McManus A. P., Gusterson B. A., Pinkerton C. R. et al. The molecular pathology of small round-cell tumors±relevance to diagnosis, prognosis, and classification. J Pathol 1996; 178: 116±121. 67. Katz R. L., Quezado M., Senderowtcz A. M. et al. An intra-abdominal small round cell neoplasm with features of primitive neuroectodermal and desmoplastic round cell tumor and a EWS/FLI1 fusion transcript. Hum Pathol 1997; 28: 502±509. 68. Fukunaga M., Nomura K., Endo Y. et al. Carcinosarcoma of the uterus with extensive neuroectodermal differentiation. Histopathol 1996; 29: 565±570. 69. Ladanyi M. The emerging molecular genetics of sarcoma translocations. Diagn Mol Pathol 1995; 4: 162±173. 70. May W. A., Lessnick S. L., Braun B. S. et al. The Ewing's sarcoma EWS/FLI-1 fusion gene encodes a more potent transcriptional activator and is a more powerful transforming gene than FLI-1. Mol Cell Biol 1993; 13: 7393±7398. 71. May W. A., Gishizky M. L., Lessnick S. L. et al. Ewing's sarcoma 11;22 translocation produces a chimeric transcription factor that requires the DNA-binding domain encoded by FLI-1 for transformation. Proc Natl Acad Sci USA 1993; 90: 5752±5756. 72. Teitell M. A., Thompson A. D., Sorensen P. H. B. et al. EWS/ETS fusion genes induce epithelial and neuroectodermal differentiation in NIH 3T3 fibroblasts. Lab Invest 1999; 79: 1535±1543. 73. Sorensen P. H. B., Shimada H., Liu X. F. et al. Biphenotypic sarcomas with myogenic and neural differentiation express the Ewing's sarcoma EWS/FLI-1 fusion gene. Cancer Res 1995; 55: 1385±1392. 74. Thorner P., Squire J., Chilton-MacNeill S. et al. Is the EWS/FLI-I fusion transcript specific for Ewing's sarcoma and peripheral primitive neuroectodermal tumor? A report of four cases showing this transcript in a wider range of tumor types. Am J Pathol 1996; 148: 1125±1138. 75. Dehner L. P. Peripheral and central primitive neuroectodermal tumors. A nosologic concept seeking a consensus. Arch Pathol Lab Med 1986; 110: 997±1005. 76. FruÈhwald M. C., O'Dorisio M. S., Fleitz J. et al. Vasoactive intestinal peptide (VIP) and VIP receptors: gene expression and growth modulation in medulloblastoma and other central primitive neuroectodermal tumors of childhood. Int J Cancer 1999; 81: 165±173. 77. Enjoji M., Hashimoto H., Tsuneyoshi M. et al. Malignant fibrous histiocytoma. A clinicopathologic study of 130 cases. Acta Pathol Jpn 1980; 30(5): 727±741. 78. Hoffman M. A., Dickersin G. R. Malignant fibrous histiocytoma. An ultrastructural study of elevan cases. Hum Pathol 1983; 14: 913±922. 79. Lawson C. W., Fisher C., Gatter K. C. An immunohistochemical study of differentiation in malignant fibrous histiocytoma. Histopathol 1987; 11: 375±393. 80. Hirose T., Kudo E., Hasegawa T. et al. Expression of intermediate filaments in malignant fibrous histiocytoma. Hum Pathol 1989; 20: 871±877. 81. Miettinen M., Soini Y. Malignant fibrous histiocytoma. Hetergeneous patterns of intermediate filament proteins by immunohistochemistry. Arch Pathol Lab Med 1989; 113: 1363±1366.

& 2001 Harcourt Publishers Ltd

665

82. Rosenberg A. E., O'Counell J. X., Dickersin G. R. et al. Expression of epithelial markers in malignant fibrous histiocytoma of the musculoskeletal system. An immunohistochemical and electron microscopic study. Hum Pathol 1993; 24: 284±293. 83. Bevis C. C. and Croker B. P. Bone and soft tissue neoplasms: Contrasts and comparison by immunoperoxidase staining for leukocyte antigens. Lab Invest 1985; 52: 7A±8A. 84. Roholl P. J. M., Kleyne J., van Unnik J. A. M. Characterization of tumor cells in malignant fibrous histiocytomas and other soft tissue tumors, in comparison with malignant histiocytes. II. Immunoperoxidase study on cryostat sections. Am J Pathol 1985; 121: 269±274. 85. Roholl P. J. M., Kleijne J., van Basten C. D. H. et al. A study to analyze the origin of tumor cells in malignant fibrous histiocytomas. A multiparametric characterization. Cancer 1985; 56: 2809±2815. 86. Wood G. S., Beckstead J. H., Turner R. R. et al. Malignant fibrous histiocytoma tumor cells resemble fibroblasts. Am J Surg Pathol 1986; 10: 323±335. 87. Brecher M. E., Franklin W. A. Absence of mononuclear phagocyte antigens in malignant fibrous histiocytoma. Am J Clin Path 1986; 86: 344±348. 88. Iwasaki H., Isayama T., Ohjimi Y. et al. Malignant fibrous histiocytoma. A tumor of facultative histiocytes showing mesenchymal differentiation in cultured cell lines. Cancer 1992; 69: 437±447. 89. Shirasuna K., Sugiyama M., Miyazaki T. Establishment and characterization of neoplastic cells from a malignant fibrous histiocytoma. A possible stem cell line. Cancer 1985; 55: 2521±2532. 90. Strauchen J. A., Dimitriu-Bona A. D. Malignant fibrous histiocytoma. Expression of monocyte/macrophage differentiation antigens detected with monoclonal antibodies. Am J Pathol 1986; 124: 303±309. 91. Binder S. W., Said J. W., Shintaku P. et al. A histiocyte-specific marker in the diagnosis of malignant fibrous histiocytoma. Use of monoclonal antibody KP-1 (CD68). Am J Clin Pathol 1992; 97: 759±763. 92. Yumoto T., Morimoto K. Experimental approach to fibrous histiocytoma. Acta Pathol Jpn 1980; 30: 767±778. 93. Ghadially F. N., Roy S. Experimentally produced synovial sarcomas. Cancer 1966; 19: 1901±1908. 94. Sakamoto K. Malignant fibrous histiocytoma induced by intra-articular injection of 9,10-dimethyl-1, 2-benzanthracene in the rat. Pathological and enzyme histochemical studies. Cancer 1986; 57: 2313±2322. 95. Katenkamp D., Neupert G. Experimental tumors with features of malignant fibrous histiocytomas. Light microscopic and electron microscopic investigations on tumors produced by cell transplantation of an established fibrosarcoma cell line. Exp Pathol 1982; 22: 11±27. 96. Krawisz S. R., Florine D. L., Scott R. E. Differentiation of fibroblast-like cells into macrophages. Cancer Res 1981; 41: 2891±2899. 97. Snover D. C., Sumner H. W., Dehner L. P. Variability of histologic pattern in recurrent soft tissue sarcomas originally diagnosed as liposarcoma. Cancer 1982; 49: 1005±1015. 98. Hashimoto H., Daimura Y., Tsuneyoshi M. et al. Soft tissue sarcoma with additional anaplastic components. A clinicopathologic and immunohistochemical study of 27 cases. Cancer 1990; 66: 1578±1589. 99. Meis J. M. `Differentiation' in bone and soft-tissue tumors. A histological indicator of tumor progression. Pathol Annual 1991; 26: 37±62.

Medical Hypotheses (2001) 57(5), 655±666

666 Zhang et al.

100. Ro J. Y., Chen J. L., Lee J. S. et al. Sarcomatoid carcinoma of the lung: Immunohistochemical and ultrastructural studies of 14 cases. Cancer 1992; 69: 376±386. 101. Nakajima M., Kasai T., Hashimoto H. et al. Sarcomatoid carcinoma of the lung. A clinicopathologic study of 37 cases. Cancer 1999; 86: 608±616. 102. Ro T. Y., Ayala A. G., Sella A. et al. Sarcomatoid renal cell carcinoma: A clinicopathologic study of 42 cases. Cancer 1987; 59: 516±526. 103. Bertoni F., Ferri C., Benati A. et al. Sarcomatoid carcinomas of the kidney. J Urol 1987; 137: 25±28. 104. Shannon R. L., Ro J. Y., Grignon D. J. et al. Sarcomatoid carcinoma of the prostate. A clinicopathologic study of 12 patients. Cancer 1992; 69: 2676±2682. 105. Wargotz E. S., Norris H. J. Metaplastic carcinoma of the breast III Carcinosarcoma. Cancer 1989; 64: 1490±1499. 106. Blom P. G., Stenwig A. E., Sobel H. J. et al. Intriguing cases. Case 6. Ultrastruct Pathol 1983; 5: 307±313. 107. Ng H. K., Poon W. S. Gliosarcoma of the posterior fosa with features of a malignant fibrous histiocytoma. Cancer 1990; 65: 1161±1166. 108. Chan J. K. C., Buchanan R., Fletcher C. D. M. Sarcomatoid variant of anaplastic large-cell Ki-1 lymphoma. Am J Pathol 1990; 14: 983±986. 109. Orndal C., Rydholm A., Willen H. et al. Cytogenetic intratumor heterogeneity in soft tissue tumors. Cancer Genet Cytogenet 1994; 78: 127±137. 110. Donner L. R. Cytogenetics of tumors of soft tissue and bone. Cancer Genet Cytogenet 1994; 78: 115±126. 111. Mandahl N., Heim S., Willen H. et al. Characteristic karyotypic anomalies identify subtypes of malignant fibrous histiocytoma. Genes Chromosomes Cancer 1989; 1: 9±14. 112. Mandahl N., Heim S., Arheden K. et al. Rings, dicentrics, and telomeric association in histiocytomas. Cancer Genet Cytogenet 1988; 30: 23±33. 113. Orndal C., Mandahl N., Rydholm A. et al. Supernumerary ring chromosomes in five bone and soft tissue tumors of low or borderline malignancy. Cancer Genet Cytogenet 1992; 60: 170±175.

Medical Hypotheses (2001) 57(5), 655±666

114. Fletcher C. D. M., Akerman M., Cin P. D. et al. Correlation between clinicopathological features and karyotype in lipomatous tumors. A report of 178 cases from the chromosomes and morphology (CHAMP) collaborative study group. Am J Pathol 1996; 148: 623±629. 115. Rosai J., Akerman M., Cin P. D. et al. Combined morphologic and karyotypic study of 59 atypical lipomaatous tumors. Evaluation of their relationship and differential diagnosis with other adipose tissue tumors. (A report of the CHAMP study group). Am J Surg Pathol 1996; 20: 1182±1189. 116. Fletcher C. D. M., Cin P. D., de Wever I. et al. Correlation between clinicopathological features and karyotype in spindle cell sarcomas. A report of 130 cases from the CHAMP study group. Am J Pathol 1999; 154: 1841±1847. 117. Ishikawa F: Breakthroughs and views. Telomer crisis, the driving force in cancer cell evolution. Biochem Biophys Res Commun 1997; 230: 1±6. 118. Genberg M., Mark J., Hakalius L. et al. Origin and relationship between different cell types in malignant fibrous histiocytoma. Am J Pathol 1989; 135: 1185±1196. 119. Antonescu C. R., Kawai A., Leung D. H. et al. Strong association of SYT-SSX fusion type and morphologic epithelial differentiation in synovial sarcoma. Diagn Mol Pathol 2000; 9: 1±8. 120. Lengauer C., Kinzler K. W., Volgelstein B. Genetic instabilities in humans. Nature 1998; 396: 643±649. 121. Pilotti S., Torre G. D., Lavarino C. et al. Distinct mdm2/p53 expression patterns in liposarcoma subgroups_implications for different pathogenetic mechanisms. J Pathol 1997; 181: 14±24. 122. Hisaoka M., Morimitsu Y., Hashimoto H. et al. Retroperitoneal liposarcoma with combined well-differentiated and myxoid malignant fibrous histiocytoma-like myxoid areas. Am J Surg Pathol 1999; 23: 1480±1492. 123. Damjanov I. Differentiation and transdifferentiation of normal and neoplastic cells. Int J Dev Biol 1996; Suppl: 1: 63S.

& 2001 Harcourt Publishers Ltd