Molecular Pathology of Soft Tissue and Bone Tumors

A Review

Tumors of the soft tissue and bone represent a heterogeneous group of mesenchymal lesions, which account for approximately 1% of all malignancies. This group includes a wide range of different tumor types with very different clinical courses and treatment sensitivities. The pathogenesis of these tumors is based on various inherited and environmental factors. This review summarizes recent advances in molecular biology that have linked the functions of tumor suppressor genes, oncogenes, and growth factors and their receptors with the behavior of soft tissue and bone tumors. Chromosome analyses, molecular cytogenetics (fluorescent in situ hybridization [FISH]), and molecular assays may become increasingly useful in diagnosis, providing important prognostic or therapeutic data and yielding new insights and approaches into the classification and treatment of these tumors. For example, the variety of tumor-specified cytogenetic and genetic alterations among small round cell tumors (Ewing/peripheral neuroectodermal family of tumors, rhabdomyosarcoma, neuroblastoma, and lymphoma) increases the possibility of genotypic diagnosis, and FISH analysis using painting probes of interphase nuclei can allow identification of tumor-specific chromosome changes in those sarcomas presenting with a low mitotic index or with poor-quality chromosomes. Hopefully, this new knowledge of genetic events will guide us toward more rational and successful development of new therapies for soft tissue sarcomas.1–5

A pathogenic role for genetic determinants has already been defined in the promotion and development of some bone and soft tissue tumors.1,2 For example, the Li-Fraumeni syndrome, characterized by germ cell mutation of the p53 tumor suppressor gene, predisposes individuals to the development of multiple tumors, including cancers of the breast and adrenal cortex, leukemia and brain tumors, and bone and soft tissue sarcomas.6 Patients with familial retinoblastoma have a higher incidence of bone and soft tissue sarcomas. Familial osteochondromas and fibrous dysplasia can also be complicated by osteosarcomas. In addition, several chromosomal aberrations and genetic mutations are strongly associated with well-defined types of tumors. It has even been suggested that racial and associated genetic characteristics may underlie differences in sarcoma susceptibility,7,8 as represented by the comparatively low incidence of sarcoma in Asians and in black Americans.8

Several chemicals can produce sarcomas; however, susceptibility to the tumorigenic effects of such chemicals is sometimes under genetic control. Ionizing radiation represents a prototype carcinogenic agent and was first recognized for the induction of osteosarcomas.9 Other soft tissue tumors commonly appearing after radiation include malignant fibrous histiocytoma (MFH), extraskeletal tumors of bone and cartilage, fibrosarcoma, hemangiosarcoma, and neurofibrosarcoma. Sarcomas of soft tissue and bone may also develop during immunosuppression, which can be associated with hematologic diseases, genetic disorders, or iatrogenic (pharmacologic) treatments to prevent graft rejection after organ transplantation. Moreover, preexisting bone infarcts, chronic osteomyelitis, and Paget disease are all associated with an increased incidence of bone tumors.

To summarize, the etiology of bone and soft tissue tumors is multifactorial. Some tumors have genetic determinants, while environmental agents and preexisting conditions are also important etiologic factors. Furthermore, in many soft tissue and bone tumors, both genetic and environmental factors interact to play a synergistic causative role.

Table 1 provides a list of defective tumor suppressor genes implicated in the development of some soft tissue and bone tumors.

The p53 gene is located at 17p13.1.10 It is composed of 11 exons, and translation of the p53 mRNA generates a protein consisting of 393 amino acids, which after phosphorylation has a molecular weight of approximately 53 kd.10,11 The p53 protein acts as a nuclear transcription factor that inhibits cell proliferation by activation of the gene encoding p21 protein, which causes some human cells to arrest at the G₁ stage of the cell cycle.10–13 These functions are regulated predominantly by N terminal encoded by exons 2–4.10,11 The p53 protein also binds to proteins involved in DNA repair and transcription.12 The C terminus encoded by exons 9–11 mediates p53 oligomerization of the protein and is involved in recognition of DNA damage and induction of apoptosis. Mutations generating defective p53 may represent early steps of carcinogenesis in bone and soft tissue or determine behavior of the developing tumor. For example, mutations of p53 have been detected in MFH, liposarcomas, leiomyosarcomas, angiosarcomas, fibrosarcomas, synovial sarcomas, Ewing sarcoma, rhabdomyosarcomas, chondrosarcomas, osteosarcomas, and other neoplasms of bone.8,13–21 It has been further suggested that overexpression of mutant p53 is associated with a less-differentiated phenotype and more aggressive behavior in soft tissue and bone tumors.13,17–21 The immunocytochemically detected overexpression of p53 protein is assumed to be the result of gene mutations that result in a mutant protein with a prolonged half-life. Nuclear p53 overexpression is a common event in MFH and has no prognostic implications in this setting,16 whereas overexpression in other soft tissue tumors may serve as a useful screening method for p53 gene mutations.15 In the evaluation of prognosis, however, the specific type of mutation, not the general p53 mutation status, should be applied.14

The retinoblastoma (Rb) tumor suppressor gene is located on chromosome 13q14 and encodes a 105-kd nuclear phosphoprotein, which plays an important role in the regulation of cell proliferation at the G₁ checkpoint.11 Although the Rb protein is expressed throughout the cell cycle, the level of its phosphorylation is cell cycle–dependent and is a target for the enzymatic activity of cyclindependent protein kinases (Cdk) complexes. Nonphosphorylated or underphosphorylated Rb protein binds to transcription regulators, such as E2F, thus preventing the cells from progressing through the G1 and S phases.22,23 Mutations at the Rb locus or genetic alterations that lead to production of malfunctioning Rb protein have been detected in osteosarcoma, MFH, liposarcoma, leiomyosarcoma, fibrosarcoma, and spindle cell sarcoma.13,24,25 It has also been suggested that alterations in the Rb protein may be the primary event during sarcoma development. Although Rb protein alterations appear to be involved in the early steps of carcinogenesis, the complexity of the Rb gene and its product, as well as the random pattern of its point mutations (ie, the absence of a well-defined “hot spot”), prevents us from fully understanding their role in tumor formation.11 Nevertheless, immunocytochemical analysis of Rb expression could serve as a screening step for more specific analysis of molecular alterations of the Rb gene.

Additional negative cell cycle regulators, which appear to serve as tumor suppressor genes involved in development of bone and soft tissue tumors, include the p21(WAF1/CIP1),p16(MTS1/INK4A), and p18 genes.11,13,25–30 The p21(WAF1/CIP1) gene maps to chromosome 6p21.1, its expression is controlled by p53, and it encodes an 18-kd protein that inhibits Cdk activity.11,13,26 The p16 gene maps to chromosome 9p21 and encodes a protein of approximately 16 kd, which by interacting with Cdk4 and Cdk6 inhibits Rb phosphorylation and induces the G₁ arrest.11,28,29 The mutation at the p16 locus has been detected in osteosarcomas, leiomyosarcomas, and some other soft tissue sarcomas.13,25,28 The p18 gene maps to chromosome 1p32,¹¹ a region also altered in leiomyosarcoma.30 The p18 gene product inhibits progression through the G₁/S phase by inhibiting kinase activity of cyclin D-Cdk4 and -Cdk6 complexes.11 The specific role of p21, p16, and p18 genes in the development of bone and soft tissue tumors requires further testing.

Overexpression of several oncogenes has been reported in tumors of soft tissue and bone (Table 2).31–37

The genes coding for nuclear transcription factors include myc, myb, gli, and fos.31–35 The c-myc and c-myb are nuclear phosphoproteins that stimulate cell proliferation (c-myc) or inhibit cell differentiation (c-myb) by binding to specific DNA sequences.38 Myc protein has a transcriptional activation domain, a DNA binding domain, a nuclear localization signal, a helix-loop-helix motif, and a leucine-zipper motif, which allows for the formation of dimers necessary for transcriptional activity.38 Amplification, alteration, or increased expression of c-myc has been detected in osteosarcomas, soft tissue sarcomas, and MFH.31–33,35 Increased expression of c-myb was found in some soft tissue sarcomas,35,38 while c-fos DNA amplification was detected in liposarcomas but not in other types of soft tissue tumors.35 Amplification of the gli gene was found in childhood sarcomas.34

The ras family is a second group of oncogenes showing alteration or increased expression in sarcomas.31,35,36 It includes H-, K-, and N-ras, which encode membrane-bound proteins associated with GTP (guanosine triphosphate)-binding/GTPase activity.38 Mutations or changes in expression of ras genes were detected in MFH; embryonal, alveolar, and pleomorphic rhabdomyosarcomas; leiomyosarcoma; and in angiosarcomas induced by vinyl chloride.35,36,38 Although alterations or changes in nuclear oncogenes or in members of the ras family have been found in many soft tissue tumors, their random pattern requires further studies on the use of this group of oncogenes as clinically significant markers of tumor progression. A potential role of deregulated G-protein–dependent pathways is emphasized by recent studies showing that activation of Gsα in osteoblastic cells results in overproduction of a disorganized fibrotic bone matrix in polyostotic and monostotic fibrous dysplasia.39

Oncogenes coding for growth factors or their receptors, or for nonreceptor tyrosine kinases, might correlate in some tumors with increased malignant potential.35,37,40,41 For example, in osteosarcomas, an increased expression of Erb-B2 appears to correlate strongly with high metastatic potential and poor prognosis.37 The coding gene, c-erb-B2 (HER2/neu) is located on chromosome 17q21 and encodes a 185-kd glycoprotein expressing tyrosine kinase activity, which shares a high homology with epidermal growth factor receptor (EGF-R).37,38,40 A large fraction of osteosarcomas, approximately 42%, express the Erb-B2 protein and its expression correlates with malignant behavior, although no significant alteration in the erb-B2 gene was found in those tumors.37 The Met proto-oncogene product is also a membrane-bound tyrosine kinase that acts as receptor for hepatocyte growth factor (HGF).38 An increased production of Met mRNA and its translation was found in sarcoma lines and soft tissue sarcomas.42 Of interest, increased expression of met was accompanied by the expression of HGF, suggestive of paracrine or autocrine regulatory mechanisms.42 Furthermore, scatter factor (SF), which is identical to HGF, can play a role in the initiation and maintenance of Kaposi sarcoma.43 In Ewing sarcoma and peripheral neuroectodermal tumors (PNET), expression of c-kit receptor and its activation by stem cell factor prevents apoptosis of neoplastic cells.44 Occasionally, sarcomas may express oncogenes classified as receptor tyrosine kinases, such as fms (mutant colony-stimulating factor 1 receptor), axl, and tie.40

Nonreceptor tyrosine kinases form another group of oncogenes with increased expression in some soft tissue tumors, including the membrane-associated nonreceptor tyrosine src family.38 This family comprises at least 9 members (yes, fgr, lyn, lck, fyn, hck, yrk, crk, and src) and, complemented with alternate translation initiation codons and tissue-specific alternative mRNA splicing, can generate more than 14 different Src type proteins.38 Common to these proteins are their tyrosine kinase activity, the presence of N-terminal myristorylation signal necessary for association with cell membrane, SH2 and SH3 domains, a kinase domain, and a C-terminal regulatory tail. Nonreceptor tyrosine kinases, such as abl and fes, have been detected in human sarcomas.40 Fes protein has been found in some low-grade tumors, and Abl protein is expressed in angiomyolipoma, in low-grade MFH, and at a low level in leiomyosarcomas. In contrast, the expression of tyrosine kinases involved in cell adhesion (HGK1/FAK) was found in angiomyolipomas, liposarcomas, leiomyosarcomas, neurofibrosarcomas, synovial sarcomas, and MFH.40 High expression of FAK was observed in the myogenic tumors and low-grade MFH.40

Other genes include sarcoma-amplified sequence (SAS) and murine double minute 2 (MDM2) genes.1,2,45–49 The SAS gene is located at chromosome 12q13–14,42 and the SAS gene product is a member of the transmembrane 4 superfamily of proteins, which regulate growth-related cellular processes.45 SAS gene amplification has been documented in both MFH and liposarcomas.42–45 The MDM2 protein is believed to inhibit wild-type p53 protein activity designed to arrest cells with damaged DNA.11,42,49 The MDM2 gene product is a 90-kd zinc-finger protein that contains a p53-binding site.11,49 The MDM2 gene maps to chromosome 12q13–14.11,42,49 It has been proposed that MDM2 is involved in the regulation of cell cycle via an autoregulatory feedback loop. The p53 protein regulates MDM2 gene transcription, and by binding to p53, MDM2 protein inhibits wild-type p53 protein transcriptional activity. Amplification of the MDM2 gene was detected in a large panel of soft tissue sarcomas and in metastatic osteosarcoma.11,42,47–49 Increased production of MDM2 mRNA and protein has been observed also in several soft tissue sarcomas.42

Multidrug resistance (MDR) can profoundly affect chemotherapy of malignant tumors. Multidrug resistance is characterized by reduced drug accumulation inside cells due to overexpression of the MDR1 P-glycoprotein, which acts as a transmembrane energy-dependent drug-efflux pump.50,51 MDR1 product expression has been detected in several bone and soft tissue tumors, such as osteosarcoma, chondrosarcoma, rhabdomyosarcoma, MFH, leiomyosarcoma, liposarcoma, undifferentiated sarcoma, synovial sarcoma, Ewing sarcoma, and neurogenic sarcoma.52 Expression of MDR1 at both RNA and protein levels could be stimulated by chemotherapy, and a positive correlation between expression of the MDR1 gene and an unfavorable prognosis was found.52 Therefore, it was proposed that MDR1 expression could serve as a molecular marker of prognosis and response to chemotherapy in soft tissue and bone tumors. Since increased expression of MDR1 may accompany tumor progression, some authors have suggested that its overexpression may also serve as a molecular marker of tumor progression.52 In the case of osteosarcoma, the MDR phenotype has been associated with unfavorable clinical outcome.53 However, recent data suggest that this phenomenon is not related to increased metastatic potential of osteosarcoma, but rather is restricted to the lack of responsiveness to cytotoxic drugs by the cells overexpressing P-glycoprotein.54

Proliferation and differentiation of mesenchyme-derived cells are coordinated by peptide growth factors that activate corresponding receptors expressing tyrosine kinase activity in paracrine, autocrine, and intracrine manners.40,55–59 Autocrine regulation occurs when the same cell produces growth factor and responds to it. In the paracrine model, one cell type produces the growth factor, while the second type responds to it. The intracrine model represents intracellular production of the factor and activation of the receptor without leaving the cell. Within tumors, growth factors can be produced by malignant cells themselves; by stromal elements, including fibroblasts, endothelial cells, and immune cells; or they can be released from carrier circulating cells, such as platelets. Since both malignant and normal cellular components of the tumor express growth factors and corresponding receptors, a complex regulatory network is formed. It includes direct paracrine or autocrine regulation and an indirect paracrine network. The latter is formed by a cascade of growth factors produced by normal or abnormal cellular elements of the tumor, which directly or indirectly activate malignant cells and the supporting mesenchymal lattice.

The growth factors and their receptors that are implied in the regulation of the phenotype of bone and soft tissue tumors are listed in Table 3. They include platelet-derived growth factor (PDGF), epidermal growth factor (EGF), transforming growth factor (TGF)-α and -β, fibroblast growth factor (FGF), insulin, insulin-like growth factor (IGF), vascular endothelial growth factor (VEGF), and HGF/SF.40,55–61 The growth factor receptors form at least 14 different receptor families, containing more than 50 receptor tyrosine kinases.55 The common traits of these receptors are presence of an extracellular ligand-binding domain, a single transmembrane region, a cytoplasmic portion with a conserved protein tyrosine kinase domain, and regulatory sequences that are subject to autophosphorylation or phosphorylation and dephosphorylation by exogenous kinases and phosphatases.38,56,58 Regulation of signal transduction by receptor tyrosine kinase involves ligand-induced receptor oligomerization via homodimerization or heterodimerization.56 Heterodimerization of the extracellular domains can increase the diversity of ligands recognized by individual receptors, while heterodimerization of the cytoplasmic portion increases the spectrum of signaling pathways because of possible recruitment of SH2 (Src-homology 2) domain-containing signaling molecules.56 Both receptor activity and density are downregulated by ligand binding, the latter through reduction of its cell surface expression via ligand-induced endocytosis.38,57 The endocytic vesicles form endosomes in which decreasing pH induces ligand dissociation. After fusion of endosomes the CURL (compartment of uncoupling of receptor and ligand) is formed, and ligand accumulating in its vesicular portion is delivered to lysosomes for degradation. The spared receptor structures from the tubular portion of CURL are recycled to the surface of the cells.

The role of PDGF in the growth of soft tissue and bone tumors is well recognized.58 Platelet-derived growth factor can be transported to the tumor site by circulating platelets or can be produced locally by endothelium or immune cells in the tumor. Platelet-derived growth factor consists of a group of disulfide-bonded homodimers and heterodimers of A and B chains with a molecular weight of approximately 30 kd (PDGF-AA, PDGF-BB, and PDGF-AB).38,58 The A and B chains show 60% of amino acid homology and are encoded by separate genes located on chromosomes 7 and 22, respectively.1,2,38,58 During tumoral transformation, the requirement for PDGF diminishes significantly, because in some systems the tumor cells can produce PDGF-like molecules.38 For example, fibrosarcomas, osteosarcomas, and glioblastomas produce PDGF-like factors. Furthermore, the transforming protein v-sis of the simian sarcoma virus is highly homologous to the PDGF-B (c-sis product).38 In addition, it has been reported that PDGF plays an important role in the proliferation of osteosarcoma cells, which can be inhibited by specific antibodies against PDGF.38,58,59 The mitogenic signal of the PDGF is translated into a phenotypic effect via the PDGF receptor (PDGF-R). There are 2 forms of PDGF-R, α and β.38 The α form has a molecular weight of 170 kd and binds the PDGF isoforms AA, BB, and AB with the same affinity. The β form has a molecular weight of 180 kd and binds PDGF-BB with high affinity and PDGF-AB with lower affinity. The α and β receptor forms are structurally and functionally related. Their extracellular ligand-binding region is made of 5 immunoglobulin-like domains and contains a single transmembrane segment with 2 cytoplasmic tyrosine kinase domains with an intervening sequence and autophosphorylation sites.38 Binding of PDGF to PDGF-R activates tyrosine kinase activity, which is optimized by receptor oligomerization.38,56 By interacting with multiple signal-transducing systems, including phospholipase A and C, phosphoinositol kinase, RasGAP, pp60^c-src, p62^c-yes, p59^fyn, Nck, and GRB2, the PDGF-R can initiate a mitogenic cascade.38 Expression of tyrosine kinase genes appears to correlate closely with expression of PDGF-R in soft tissue tumors, such as angiolipoma, liposarcoma, leiomyosarcoma, synovial sarcoma, and MHF.62 However, this expression was heterogeneous. For example, overexpression of β-PDGF-R mRNA is not related to grade or specific for cell type, in agreement with immunocytochemical studies showing variable expression of this receptor in sarcomas.62 Nevertheless, high levels of α-PDGF-R expression have been observed in metastatic or high-grade sarcomas.62 Therefore, although local PDGF/PDGF-R loops could play an important pathogenic role in the growth of sarcomas, the clinical use of PDGF or PDGF-R expression for diagnostic purposes is limited.

The FGF family is composed of 9 members with an approximate molecular weight of 20 to 30 kd (Table 3). They play an important role both in the regulation of growth of normal and malignant cells of mesenchymal origin and in stimulating angiogenesis.38,55,57 FGF-1 and FGF-2 represent acidic and basic FGF, respectively. FGF-4 plays an important role in the development of Kaposi sarcoma by stimulating angiogenesis. A common feature shared by all FGFs is their high binding affinity for heparan sulfate and heparan sulfate proteoglycans.38 The FGF receptor (FGF-R) family is composed of 5 membrane-bound tyrosine kinases.38,63 They share significant sequence homology and overlap in their binding specificities for different FGFs. The common structure of the FGF-R involves an extracellular segment composed of 3 disulfide-bonded loops, an immunoglobulin-like domain, a single transmembrane segment followed by an juxtamembrane domain that is 80 amino acids long, 2 tyrosine kinases separated by a 14-amino acid insert, and a C-terminal tail tyrosine autophosphorylation site.

Insulin, in addition to its well-known metabolic effects, can act as a growth factor directly or synergistically with other growth factors.64 The insulin signal is translated into phenotypic effect by the insulin receptor, which is a disulfide-linked tetramer composed of 2 α subunits of 125 kd and 2 β subunits of 90 kd. The insulin receptor has serine, threonine, and tyrosine kinase activities and can autophosphorylate the tyrosine residues on the C terminal of the β subunit.64

Insulin-like growth factor I and IGF-II are related proteins with a molecular weight of approximately 7.5 kd.38,57,59,65 Originally, they were described as serum factors that were able to stimulate cartilage growth in vitro.59 The IGF-I gene is located on chromosome 12 and contains 5 exons, while the gene for IGF-II is located on chromosome 11 and consists of 9 exons.65 The IGF-I receptor is structurally similar to the insulin receptor and is composed of 2 α subunits of 125 kd and 2 β subunits of 90 kd that have tyrosine kinase activity.60,66 The IGF-II receptor is a transmembrane protein of 270 kd, containing a small cytoplasmic domain that lacks kinase activity but expresses mannose 6-phosphate receptor activity.38,64 Insulin and IGFs can cross-interact with corresponding receptors, although with lower affinity than the native ligand.64–66 Insulin-like growth factor I is produced by the liver and peripheral tissues, including tumor stroma and malignant cells.38,65 Similarly, IGF-II is also produced by both normal and neoplastic tissues. Insulin-like growth factors can play an important role in the growth of soft tissue and bone tumors, as IGF-I can be produced and interact with its corresponding receptor in Ewing sarcoma, PNET, and osteosarcomas.55,57,59,67,68 The significance of this effect can be expressed by actual inhibition of tumor growth with monoclonal antibodies and antisense nucleotides against the IGF-I receptor, as seen in some osteosarcomas,67 or inhibition of the IGF-I/IGF-I receptor pathway, as in Ewing sarcoma/PNET.68 Insulin-like growth factor II can be produced by rhabdomyosarcoma and leiomyosarcoma and may play a role in the malignant behavior of these tumors in addition to producing severe hypoglycemia.59,69

The EGF family includes EGF and TGF-α.38,57 Epidermal growth factor is a peptide growth factor of approximately 6 kd, which is synthesized as part of a large protein that is processed to mature EGF and EGF-like polypeptides. Transforming growth factor-α is also produced as a 160-amino acid precursor peptide that is cleaved to a biologically active, 50-amino acid peptide. Both TGF-α and EGF can regulate proliferation of mesenchymal and epithelial cells. Their signal is transduced through interaction with the EGF-R, a monomeric transmembrane glycoprotein of about 170 kd that has tyrosine kinase activity. The C domain of the EGF-R contains 4 sites for tyrosine autophosphorylation and several sites for serine/threonine phosphorylation. A potential role for EGF-R in the growth of mesenchymal tumors is suggested by the finding that expression of Erb-B2, which shares high homology with EGF-R, correlates with a poor prognosis in osteosarcoma patients.37

The TGFβ family consists of 5 TGF-β peptides (β1–β5) of approximately 25 kd that participate in growth and differentiation processes, such as bone formation, extracellular matrix formation, angiogenesis, and development of chemotaxis in fibroblasts; they also act as immunosuppressors.38,70 In cell culture, the TGF-βs predominantly stimulate proliferation of mesenchymal cells, while inhibiting growth of epithelial cells; in vivo, they act mostly as growth inhibitors. There are 2 TGF-β receptors, I and II, which are composed of 53- and 75-kd transmembrane glycoproteins, respectively, with serine/threonine kinase activity.71 The membrane-bound TGF-β–binding glycoproteins include TGF-β–binding protein type III (betaglycan; a protein of 280 kd), type IV (64 kd), type V (400 kd), and type VI, and phosphoinositol-anchored proteins of 60, 140, and 180 kd.38,70,71 These proteins can bind the different forms of TGF-β with different affinities. The information conveyed by TGF-βs is translated into phenotypic effect following binding to the extracellular binding site, receptor oligomerization, and activation of cytoplasmic serine/threonine kinase activity of receptors I and II, or by interaction with other membrane-associated TGF-β–binding proteins.38,71

The VEGF/VPFs and SF/HGFs play a crucial role in normal tissue and tumor angiogenesis.60,61,72 The degree of vascularization in a tumor has a profound effect on its growth and behavior. The VEGF is a dimer of two 20- to 23-kd subunits and consists of at least 4 subtypes.60 Vascular endothelial growth factor stimulates growth of endothelial cells and increases vascular permeability via interaction with the VEGF receptor, a product of the flt-1 gene.60 This gene is 1 of 3 members of the flt family of receptor transmembrane tyrosine kinases located on chromosome 13q12-q13 that share high sequence homology.60 The VEGF-flt receptors play an important role in the development and maintenance of the circulation system and normal tissue vascularization, as well as in tumor growth and development. The SF/HGF is a mesenchyme-derived protein composed of a 60-kd α chain and a 30-kd β chain.61 The SF/HGF receptor has a molecular weight of 190 kd and is the product of the c-met proto-oncogene; it is composed of a 145-kd transmembrane spanning β chain and a 50-kd α chain expressed on cell surface.61 The β chain contains an extracellular binding site for SF, intracellular tyrosine kinase activity, and phosphorylation sites.61 Via interaction with the c-met receptor, SF regulates angiogenesis in normal and malignant tissues. The SF is produced predominantly by mesenchymal cells, such as fibroblasts, vascular smooth cells, glial cells, macrophages, and T lymphocytes.61 It has been proposed that activation of the SF-c-met receptor system is important for development of Kaposi sarcoma.43,61 The important role of vascularization in tumor development is emphasized by a potent antitumor effect of angiogenesis inhibitors, including angiostatin and endostatin.72,73

Deregulation or decreased expression of additional factors, such as growth-suppressing gene (GOK)74 and integrin α6 subunit (VLA-6),75 may affect growth of soft tissue tumors. GOK was identified at 11p15.5, and it is a transmembrane protein that may be a receptor connected to pathways that inhibit growth in rhabdomyosarcoma and rhabdoid tumors.74 VLA-6 integrin serves as a major adhesion receptor for laminin and may be an important factor in the invasiveness and metastatic potential of fibrosarcoma.75 Recently, it was proposed that inactivation of normal alleles for the multiple-pass patched transmembrane receptor (ptc) derepresses ptc with subsequent activation of gli-1 and upregulation of IGF-II, which in turn stimulate rhabdomyosarcoma development.76

Autocrine or paracrine loops between c-kit and stem cell growth factor have been proposed as important regulators of behavior in sarcomas of neuroectodermal origin, by which stem cell growth factor, through an interaction with c-kit, can inhibit tumor cell apoptosis.44

DNA studies utilizing flow cytometry, image analysis, and cytogenetics have been shown to be of prognostic value in many tumors of the soft tissue and bone.1,2,77,78 In general, aneuploidy and increased ploidy tend to reflect changes in chromosome number and may therefore correlate with increased malignant behavior. However, it must be emphasized that early-stage lesions can be aneuploid,1,2 and that aneuploidy or increased ploidy in some soft tissue tumors does not necessarily correlate with their malignant potential.79

Benign soft tissue tumors are usually diploid, with the exception of juvenile angiofibroma.1,2 Flow cytometric analyses of childhood inflammatory myofibroblastic tumors (inflammatory tumors) have shown that although both diploid and hyperdiploid patterns may be detected, only samples from patients with clinical recurrence are hyperdiploid.79 These findings are in agreement with retrospective studies showing that, at least in the bladder, inflammatory tumors are diploid.80 Comparison between different soft tissue sarcomas has shown a significant correlation between tumor grade and frequency of aneuploidy for high-grade (stages III and IV) sarcomas. Aneuploidy carries a less favorable prognosis in MFH, synovial sarcoma, and clear cell tumor. Unexpectedly, childhood rhabdomyosarcoma patients with aneuploid tumors (hyperdiploid DNA content) have shown significantly increased survival rates, whereas the relationship between malignant behavior and DNA ploidy status does not hold true for epithelioid sarcoma, liposarcoma, angiosarcoma, malignant rhabdoid tumor, alveolar soft part sarcoma, and hemangiopericytoma. Therefore, for diagnostic purposes, testing of DNA ploidy should be combined with karyotype analysis or molecular probing of the DNA or RNA samples.

Cytogenetic analysis has shown a remarkably high incidence of tumor-specific chromosomal aberrations in bone and soft tissue tumors.1–5,77,78 Such aberrations include general changes in chromosome number or specific chromosomal rearrangements, such as translocations, deletions, insertions, inversions, duplications, or gene amplifications. The most characteristic chromosomal aberrations are presented in Table 4.

In Ewing sarcoma and PNET tumors, including peripheral neuroepithelioma, Askin tumor, and esthesioneuroblastoma, the most common and characteristic chromosomal abnormality is the translocation t(11;22)(q24;q12).1,2,81 Nevertheless, this translocation has also been detected in single cases of small cell osteosarcoma and mesenchymal sarcoma, while some PNET tumors do not contain t(11;22).1,2 The second most common abnormality in Ewing sarcoma is a trisomy 8 and der(16)t(1;16)(q21;q13).2 The translocation of chromosomes 11 and 22 is also characteristic for desmoplastic small round cell tumor, although the breakpoint on chromosome 11 for this translocation t(11;22)(p13;q11–12) is on a different arm than the corresponding defect in Ewing sarcoma.1,2 Many neoplasms that are of different origin share similar morphologic features that have been defined as small round blue cell tumors; these neoplasms include various types of PNET (Figure 1), Ewing sarcoma, some forms of rhabdomyosarcoma and mesenchymal chondrosarcoma, small cell osteosarcoma, hemangiopericytoma, desmoplastic small round cell tumor (Figure 2), and neuroblastoma. Therefore, since many of these tumors have characteristic karyotypic features, chromosome analysis can provide important information supporting phenotype-based diagnosis.

Rhabdomyosarcomas are also characterized by cytogenetic abnormalities, which have a different nature in the alveolar form than in the embryonal form.1–4,82 Thus, the characteristic rearrangement in alveolar rhabdomyosarcoma (Figure 3) is t(2;13)(q37;q14), although variants have translocations involving 13q14, q35, and q36.1.1 In addition, t(1;13) and t(2;11) have been reported.83 Only in exceptional cases have these translocations been found in nonalveolar rhabdomyosarcomas.1 It is important to emphasize that detection of the t(2;13) and t(1;13) translocations is readily feasible using the reverse-transcription polymerase chain reaction (RT-PCR) technique or by FISH analysis of small formalin-fixed biopsies.83

Embryonal rhabdomyosarcomas are, in turn, characterized by more complex numerical and structural abnormalities, including trisomy 8 and changes in chromosomes 2 and 11.1,2,77 Loss of heterozygosity on 11p in embryonal rhabdomyosarcomas suggests that it may be a primary genetic event in this tumor, with the chromosome changes being secondary to loss of heterozygosity.1 Primitive or undifferentiated rhabdomyosarcomas may show all or any of the abnormalities described above, while those with mixed morphologic elements may have combinations of the abnormalities characteristic of both rhabdomyosarcoma and Ewing sarcoma. Benign tumors of striated muscle do not exhibit the abnormal karyotype.2 The smooth muscle tumors, leiomyomas, consistently contain chromosomal aberrations that include structural and numerical changes in chromosomes 6, 7, 12, and 14.1,2,15,77 In leiomyosarcomas, chromosomal abnormalities are heterogeneous. Sandberg and Bridge1 proposed categorizing the tumors into 3 cytogenetic subgroups. The first group is characterized by a pseudodiploid chromosome pattern resulting from reciprocal translocations; the second subgroup is characterized by hypodiploidy with consistent monosomy 22, monosomy 18, and/or partial monosomy 1p13-pter1; and the third subgroup encompasses the remainder of the cases, which fail to display any consistent chromosome pattern.

Lipomas represent cytogenetically diverse benign tumors of fat tissue.1,2 In patients with multiple lipomas, the majority of the tumors show a normal karyotype (98%), whereas in patients with sporadic lipomas, only 22% of tumors have a normal karyotype.1 The most frequent chromosome changes involve 12q13–15, t(3;12)(q27–28;q13–15), and 6p with the most common breakpoint at 6p21–23, and 13q.1,84 Deep-seated lipomas or atypical lipomas frequently exhibit ring chromosomes.1 Rare aberrations have included 8q, found in lipoblastoma; 11q, found in hibernoma; and 16q, which is found in spindle cell/pleomorphic lipoma.84 In a study of a small number of angiolipomas, Fletcher et al demonstrated that the karyotype was normal.84 The cytogenetic diversity of lipomas in the context of their benign character suggests the involvement of proliferation-related genes, but not of putative genes responsible for malignant transformation.1

In liposarcomas, 50% of tumors have an abnormal karyotype; the most common abnormality is a simple reciprocal translocation, usually involving chromosome 12.1 A characteristic change for myxoid liposarcoma (Figure 4) is translocation t(12;16)(q13;p11), a change that has not been found in other myxoid tumors.1 Recent cytogenetic studies in round cell liposarcoma have also shown a t(12;16)(q13;p11) translocation.1 In addition, the FUS-CHOP transcript, which is indicative of t(12;16), has been detected in some well-differentiated and pleomorphic liposarcomas.85 Additional abnormalities found in myxoid liposarcoma involve chromosomes 1 and 8. Pleomorphic liposarcomas show multiple and complex structural chromosomal changes and translocations. Although similar absence of a characteristic cytogenetic pattern has been found in well-differentiated liposarcoma, the abnormalities observed did not overlap with those of pleomorphic liposarcoma. The most characteristic features of well-differentiated liposarcomas were the presence of ring chromosome, large marker chromosomes, and telomeric associations.1

Among benign tumors of the connective tissue, desmoplastic fibroma revealed abnormalities of chromosomes 3 and 5.1 Desmoid tumors have characteristically shown loss of chromosome Y and interstitial deletions of the long arm of chromosome 5.86 Fibrous tissue from areas affected by Dupuytren contracture and Peyronie disease have shown trisomy of 7 or 8.1 The giant cell tumor of bone is characterized by telomeric associations, telomere-to-telomere translocations which most commonly affect telomeres from 19q, 11p, 15p, 18p, 20q, and 21p.1 Tenosynovial giant cell tumors may also have numerical and structural aberrations.1,2

Malignant fibrous histiocytoma of bone is characterized by a wide spectrum of karyotypic abnormalities, such as numeric changes ranging from haploid to hypertetraploid chromosome number, and a homogeneous staining region on the short arm of chromosome X or 6. Other structural rearrangements have involved chromosomes 1, 3, 5, 15, and 17. Common karyotypic abnormalities in MFH of soft tissue involve chromosome 19, 1q, 3p, and 11p and presence of ring chromosomes. In dermatofibrosarcoma, karyotypic changes involve chromosomes 17 and 8 and the presence of ring chromosomes. In dermatofibrosarcoma protuberans (Figure 5), Naeem et al recently demonstrated that amplification of interspersed sequences of chromosomes 17 and 22, in ring form, represents a characteristic aberration.87 Other authors documented that both supernumerary ring chromosomes derived from t(17;22) and reciprocal translocation t(17;22)(q22;q13) are characteristic for dermatofibrosarcoma and giant cell fibroblastoma.88 In fibrosarcomas, predominant cytogenetic abnormalities include numerical changes of chromosomes 11, 17, 18, 19, and 20. In addition, structural changes in chromosomes X, 7, and 22 were detected.1

The alveolar soft part sarcoma (Figure 6) is characterized by changes in chromosomes 1, 5, 13, and 17.1 The desmoplastic small round cell tumor, recognized recently as a type of primitive sarcoma, is characterized by t(11;22)(p13;q12).89 Clear cell sarcoma (malignant melanoma of soft parts) (Figure 7) is characterized by specific translocation of t(12;22)(q13;q12–13).1,2,81,90,91 In the absence of (t12;22), a defect in chromosome 22 was frequently reported.1 In addition, extra copies of chromosome 8, overexpression of 1q[i(1)(q10)], del(1)(q42), I(6)(p10), deletion of 6q, and an increased number of chromosome 7 were detected. The synovial sarcoma (Figure 8) is characterized by a nonrandom translocation, t(X;18)(p11.2;q11.2).1–4,81,92–94 Complex translocations involving chromosomes X and 18 and another autosome have also been described.1 An isolated X;18 translocation is seen more commonly in primary synovial sarcoma than in recurrent tumor.1 Chromosomes most frequently involved in secondary changes include chromosomes 12 and 1. Hemangiopericytomas have rearranged chromosomal regions 13p12–21, 11p11–13, 12q13–15, and 19q13, which appear to be nonrandom abnormalities.1,2

Little information is available on cytogenetic anomalies in osteomas, osteoid osteoma, and osteoblastomas.1,2 One fully characterized osteoblastoma exhibited the karyotype 46,XY,+der(15)t(15;20)(p11;p11),der(17)t(17;20)(p11–12)(q11),−20.2 Osteosarcoma is the most common primary tumor of the bone. Osteosarcomas have shown complex karyotypes, which are characterized primarily by multiple unbalanced structural abnormalities.1,2 The structural changes include chromosomes 1, 3, 7, 11, 17, and 22, while losses of chromosomes 3, 10, 13, and 15 represent the most common numerical anomalies. Although osteosarcomas have a very complex karyotype, tumor-specific loss of heterozygosity was most frequently reported for 3q, 13q, 17p, and 18q.93 Of note, the potential tumor suppressor locus for osteosarcoma was localized to 3q26.2–3q26.3 between the loci D3S1212 and D3S1246.93

In chondrogenic tumors, chondromas frequently have abnormalities involving 12q13–15.1,2 Analysis of chondrosarcomas showed a predominant structural aberration of chromosomes 1p36, 1p11, 1q21, 5q13, 11p15, 12q13–15, 15p11, 19p13, and 20q11. Numerical aberrations involved chromosomes 5, 7, 8, 10, 18, and 20, of which trisomy for chromosomes 7 and 20 and monosomy for chromosome 10 are the most frequent.1,95 The involvement of 12q13–15 is considered to be significant, inasmuch as it is also frequently altered in other mesenchymal neoplasms, such as lipoma, myxoid liposarcoma, and clear cell sarcoma. Extraskeletal myxoid chondrosarcoma is characterized by translocation t(9;22)(q22;q11–12), and osteocartilaginous exostoses show loss of genetic material from distal 8q.1,2

Diagnosis of sarcomas based solely on morphologic features may be difficult; therefore, the detection of tumor-specific translocations represents an additional diagnostic tool. Specific molecular defects can be defined with RT-PCR, DNA-based PCR, Southern blotting, and FISH methods.3–5,35,77,81–98 Hematopoietic and soft tissue sarcomas, but not carcinomas, characteristically display chimeric transcription factors resulting from specific translocations.1,2,81,96 This phenomenon most likely reflects the embryonic and postnatal reliance of hematopoietic and mesenchymal tissues on transcription factor networks to allow programming of heterogeneous differentiation from shared stem cells and progenitors.96 In sarcomas, some translocations and resulting chimeric transcripts have already been characterized at the molecular level and can therefore be used for classification of these tumors (Table 5).*

Ewing sarcoma and PNET (Figure 1) are characterized by the translocation t(11;22).1,81,96,98 Molecular analysis has revealed that the EWS gene on chromosome 22q12 is fused to the FLI1 gene, which is mapped to 11q24.8,81,89 The chimeric gene product of this translocation, EWS-FLI1, produces a fusion protein containing N-terminal sequences of EWS linked to the Ets-like DNA-binding domain of FLI1.38,96 The conserved box present at the C-terminal region of EWS functions as an RNA-binding domain, which specifically binds to poly G and poly U.38,96 The FLI1 protein, a member of erg family, acts as a sequence-specific transcriptional activator with 2 autonomous transcriptional activation domains, one at the N-terminal and the second on the C-terminal regions.38,81 In the protein product of the EWS-FLI1 gene, the RNA-binding domain of EWS is replaced by the DNA binding domain of FLI1, and the N-terminal domain of EWS acts as a strong transactivator.81 There are at least 9 types of chimeric EWS-FLI1, representing different combinations of exons from EWS and FLI1, although the entire DNA-binding domain of FLI1 and the entire N-terminal domain of EWS are present in all variants.81 It has been suggested that the chimeric protein could disrupt transcriptional regulatory pathways, thus being responsible for the genesis of Ewing sarcoma and PNET tumors. Recently, Thorner et al99 reported that the EWS-FLI1 transcript was expressed in 2 polyphenotypic tumors and 2 mixed embryonal and alveolar rhabdomyosarcomas. The authors suggested that this exceptional finding should be taken into consideration in the molecular diagnosis of Ewing sarcoma or PNET.99 In another translocation, t(21;22)(q22;q12), EWS is fused with ERG in at least 4 molecular variants, with the resulting protein products functioning as transcriptional activators.81 The rare translocation t(7;22)(p22;q12) involves fusion of EWS with the ETV1 gene segment.81

In malignant melanomas of the soft parts (Figure 7), sequences similar to those cited above are fused to the basic region-leucine zipper protein (bZIP) domain of ATF-1 in t(12;22).91,96 AFT-1, a member of the large family of transcription factors ATF/CREB, is the transcription factor gene encoding a leucine zipper dimerization domain and a basic DNA-binding domain.81,90 EWS-ATF-1 acts as an efficient constitutive transcriptional activator that does not require cyclic adenosine monophosphate to induce native ATF-1.90 It has been suggested that replacement of part of the N-terminal kinase regulatory domain of ATF-1 with the EWS regulatory domain results in altered DNA binding, protein-protein interaction, and acquisition of transcriptional activation properties that cause deregulated gene expression, as well as acquisition of an inability to respond to cyclic adenosine monophosphate, which may be responsible for the genesis of malignant melanomas of the soft parts.90,100,101

Desmoplastic small round cell tumor (Figure 2) is characterized by t(11;22)(p13;q12).1,89 In this translocation, the EWS gene may fuse with the Wilms tumor gene, WT1, generating a fusion gene; transcription and translation of this fusion gene produce protein, and the RNA-binding domain of EWS is replaced by the 3 C-terminal zinc fingers of the WT1 DNA-binding domain.81,96 The chimeric EWS-WT1 product includes 2 alternatively spliced forms of the zinc-finger domain of WT1, which may have different binding capacities, as in native spliced forms of WT1.96 Most recently, it was reported that the DNA-binding domains of WT1 and EWS/WT1 are functionally distinct.102 It has been suggested that loss of transcriptional suppression of WT1 target genes and multilineage differentiation in desmoplastic small round cell tumor may be associated with activation of the same target genes found in heterologous elements of Wilms tumor.81

Myxoid liposarcoma (Figure 4) is characterized by translocation t(12;16)(q13;p11), which generates a fusion gene containing the sequence for the N terminal of the FUS (TLS) and bZIP domain of CHOP.1,2,81,84,85,103,104 The FUS (TLS) gene maps at chromosome 16p11, and CHOP maps at chromosome 12q13. The CHOP gene, having a leucine zipper-type dimerization motif and DNA-binding domain, may function as an inhibitor of other transcription factors, thus regulating81 or preventing105 adipocyte differentiation. The FUS-CHOP protein functions as a transcription activator. Transcription of the FUS-CHOP fusion gene generates 2 alternatively spliced mRNAs, which differ by 276 base pairs (bp).103 Of interest, although some lipomas, leiomyomas, pleomorphic adenomas of salivary gland, MFHs, and hemangiopericytomas showed cytogenetic rearrangement of 12q13–15, CHOP was not rearranged.81 In addition, the histogenetic relationship between myxoid liposarcoma and round cell liposarcoma has been confirmed by the finding of FUS-CHOP transcripts in cases of liposarcoma composed either partly or entirely of round cell areas.81,104 However, recent studies have shown that expression of the tumor transcripts was also present in some well-differentiated and pleomorphic liposarcomas.85

In alveolar rhabdomyosarcoma (Figure 3), the characteristic translocation t(2;13)(q35;q14) disrupts the PAX3 paired box gene on chromosome 2 and juxtaposes it with transcription factor gene FAHR, mapping at chromosome 13.1,81–83,96 The PAX genes bind DNA and recognize specific target genes through their paired box and homeodomain DNA-binding regions and act as important developmental regulators.96 The PAX3-FAHR. fusion gene encodes a protein containing the DNA-binding domain of PAX3, whose C-terminal sequence has been replaced with the bisected forkhead DNA-binding sequence from FAHR.81–83,96 A variant translocation, t(1;13)(p36;q14), involves FAHR and another gene from the PAX family, PAX7.81–83,96 Both fusion proteins can act as transcription regulators.

In synovial sarcoma (Figure 8), the characteristic translocation t(X;18)(p11;q11) generates a fusion gene, which produces the chimeric SYT-SSX transcript.81,92,96,97 The SSX gene contains 2 copies, SSX1 and SSX2, which are located within distinct subregions of Xp11 and are designated OATL1 and OATL2, respectively.96 The SYT gene is located at 18q11 and contains a glutamine-proline–rich region suggestive of a transcriptional activation domain.81,96 The predicted protein encoded by the SYT-SSX transcripts contains 396 N-terminal amino acids of SYT fused with the 78 C-terminal amino acids of either SSX1 or SSX2.106 In this construct, with the exception of the 3′ region, SYT replaces the 5′ portion of the SSX homologous to the Kruppel-associated box (KRAB), which is a transcriptional repression domain found in zinc-finger transcription factors.97,106 It has been suggested that this chimeric product may be involved in transcriptional events in synovial sarcoma.97,106 Further molecular characterization of synovial sarcoma in a single case revealed the presence of an SYT-SSX2 variant with an additional 126-bp segment inserted between the usual SYT and SSX2 fusion point and encoding a protein longer by 43 amino acids.97 This event could represent an insertion, a deletion, or cryptic RNA splicing. So far, the RT-PCR analysis of synovial sarcomas having t(X;18) has confirmed the presence of the SYT-SSX transcripts in more than 90% of cases and has shown that the incidence of molecular variants of the transcript is low.97 Thus, RT-PCR could be used as a highly efficient diagnostic tool for synovial sarcoma and could be useful in detecting otherwise occult metastatic cells in the circulation and in bone marrow. Other studies have emphasized the use of nonbreakpoint DNA probes to detect t(X;18) in synovial sarcoma on paraffin-embedded slides.94

In myxoid chondrosarcoma, the recurrent translocation t(9;22)(q22;q12) has been suggested as a marker of extraskeletal myxoid chondrosarcoma.107 Such translocation is absent in skeletal myxoid chondrosarcoma and represents a rearrangement of the EWS gene at 22q12 with a novel gene at 9q22 designated CHN (or TEC).107 CHN encodes a novel orphan nuclear receptor with a zinc finger DNA-binding domain.107 This translocation can be easily detected by PCR amplification of the EWS/CHN gene fusion product.

In dermatofibrosarcoma protuberans (Figure 5) and giant cell fibroblastoma, translocation t(17;22) generates a fusion gene, designated PDGFB-COL1A1.88 It represents a product composed of an extracellular matrix protein (COL1A1) and growth factor (PDGFB).

In summary, molecular analysis of translocations should be a useful tool for the differential diagnosis of sarcoma and for the prediction of tumor behavior. The specific translocations for Ewing sarcoma/PNET tumors, t(11;22), t(21;22), and t(7;22); clear cell sarcoma, t(12;22); myxoid liposarcoma, t(12;16); desmoplastic round cell tumor, t(11;22); alveolar rhabdomyosarcoma, t(2;13) and t(1;13); synovial sarcoma, t(X;18); extraskeletal myxoid chondrosarcoma, t(9;22)(q22;q12); and dermatofibrosarcoma protuberans, t(17;22), produce fusion genes whose products can act as transcriptional regulators and which can affect the behavior of target cells. Those fusion genes include EWS-FL1, EWS-ERG, EWS-ETV1, EWS-ATF1, FUS-CHOP, PAX3-FAHR., PAX7-FAHR., EWS-WT1, SYT-SSX1, SYT-SSX2, EWS-CHN, and PDGFB-COL1A1, and their expression is tumor cell–specific. The fusion genes can be detected by Southern blotting, followed by hybridization with specific probes, and by PCR complemented with Southern blotting with hybridization to establish the specificity of the amplified messages. The tissue distribution of fusion genes can be determined by the FISH method. The expression of defective genes can be measured using the Northern blotting technique or the very sensitive RT-PCR assays. The latter can be supplemented with Southern blotting and hybridization to specific cDNA probes to confirm the specificity of the product and to determine the presence of heterogeneous variants of chimeric mRNA. Reverse-transcription PCR may also allow the earliest detection of metastatic disease, before clinical and morphologic evidence of disease has developed. A challenging area for future research is the production of monoclonal and polyclonal antibodies against the actual proteins produced by the chimeric transcripts. Such antibodies would help in the immunocytochemical diagnosis of sarcomas.