Genomic imbalances of 7p and 17q in malignant peripheral nerve sheath tumors are clinically relevant.

We investigated 31 malignant peripheral nerve sheath tumors (MPNSTs) from 23 patients by means of comparative genomic hybridization (CGH) in order to study quantitative genomic aberrations of these tumors. Twenty-one of the 23 patients revealed changes, with a mean value of 11 aberrations per sample (range 2-29). The minimal common regions of the most frequent gains were 8q23-q24.1 (12 cases), 5p14 (11 cases), and 6p22-pter, 7p15-p21, 7q32-q35, 8q21.1-q22, 8q24.2-qter, and 17q22-qter (10 cases each). Seventeen high-level amplifications were detected in eight of the 21 samples. In three cases, the high-level amplifications involved 8q24.1-qter, and in two cases each the high-level amplifications involved regions 5p14, 7p14-pter, 8q21.1-q23, and 13q32-q33. The minimal common region of frequent losses was 14q24.3-qter (five cases). The gain of 8q as a single common change in the primary tumor, the recurrence, and the metastasis from the same patient suggests that this aberration is an early change in the tumorigenesis of MPNSTs. Comparable aberrations were observed in separate tumors of the same patients affected by Recklinghausen's disease, indicating a limited number of accidental secondary changes. In sporadic MPNSTs, the most frequent gains were narrowed down predominantly to 5p, 6, 8q, and 20q, whereas in MPNSTs from patients with Recklinghausen's disease, there was most often a gain in 7q, 8q, 15q, and 17q. The occurrence of gain of both 7p15-p21 and 17q22-qter was associated with a statistically significant poor overall survival rate (P = 0.0096).     

Cytogenetics of synovial sarcoma: presentation of ten new cases and review of the literature.

Cytogenetic study of five biphasic and five monophasic synovial sarcomas revealed the specific abnormality t(X;18) (p11;q11) in eight cases and t(X;15;18) (p11;q15;q11) and t(X;7) (q11-12;q32) in one case each. Additional, secondary aberrations were present in eight of these tumors. By combining our data with information on previously published cytogenetically abnormal synovial sarcomas, we were able to evaluate 32 tumor samples from 29 patients. The modal chromosome number was pseudodiploid or near diploid in 26 of the 32 tumors. A t(X;18) was present in 21 of 29 cases (72%). Complex translocations involving chromosomes X and 18 and another autosome were present in five cases, and one displayed a t(5;18). There was no visible rearrangement of chromosome bands Xp11 or 18q11 in only 2 of the 32 synovial sarcomas. Half of the primary tumors (6 of 12) had the X;18-translocation as the sole abnormality. Of the remaining 20 specimens from recurrent or metastatic tumors (in three cases two tumors could be analyzed), only one had t(X;18) as the sole change. The secondary aberrations in cases exhibiting clonal evolution were also generally more extensive in the metastatic and recurrent than in the primary sarcomas (five additional aberrations per case, compared with two). Chromosomes 1 and 12 were the chromosomes most frequently (one fourth of the cases) involved in additional structural changes, but with several different breakpoints. No differences were identified between the karyotypic profiles of monophasic and biphasic synovial sarcomas.     

Comparison of genetic changes in primary sarcomas and their pulmonary metastases.

The aims of the present study were to compare genetic aberrations in primary sarcomas and their pulmonary metastases and to explore the pathways associated with disease spreading. The primary tumor and its subsequent pulmonary metastasis of 22 patients were analyzed by comparative genomic hybridization. All samples were obtained before the initiation of chemo- or radiotherapy. The mean total number of aberrations per tumor was 7.6 (range, 0-17) in primary tumors and 7. 5 (range, 0-19) in metastases. The mean numbers of high-level amplifications per tumor were similar (0.32 in primary tumors and 0. 36 in metastases). The frequencies of the most common aberrations were relatively similar in primary tumors and metastases: the most frequent gain affected 1q (minimal common regions 1q21-q23 in 36% of primary tumors and 1q21 in 45% of metastases). The most frequent losses were detected at 9p (9p22-pter in 32% of primary tumors and 9p21-pter in 32% of metastases), 10p (10p11.2-p12 in 41% of primary tumors and 10p11.2-pter in 32% of metastases), 11q (11q23-qter in 36% of primary tumors and 32% of metastases), and 13q (13q14-q21 in 45% of primary tumors and 50% of metastases). No aberrations specific to metastases were detected. An increase in the total number of changes during progression was a predominant feature in a majority of these paired samples. Also, the number of differences in the genetic profile outnumbered common changes in a majority of the samples. However, despite the heterogeneous and numerous changes, all pairs with aberrations in both specimens had some shared alterations in both samples. Genes Chromosomes Cancer 25:323-331, 1999.     

DNA copy number changes and evaluation of MYC, IGF1R, and FES amplification in xenografts of pancreatic adenocarcinoma

We analyzed eight samples of xenografted human pancreatic tumors and two metastases developed in mice by comparative genomic hybridization (CGH). The most recurrent changes were: gains on chromosomes 8 (8q24-qter; 7/8 cases), 15 (15q25-q26; 6/8 cases), 16 (16p in 6/8 cases; 16q in 5/8 cases), 20 (20q; 6/8 cases), and 19 (19q; 5/8 cases); and losses on chromosomes 18 (18q21; 6/8 cases), 6 (6q16-q21 and 6q24-qter; 5/8 cases each), and 9 (9p23-pter; 5/8 cases). The two metastases maintained the aberrations of the original pancreatic tumor plus gain of 11q12-q13 and 22q. Loss of heterozygosity analysis was carried out for 10p14-pter, a region that was lost in 3/8 samples. All of them presented allelic imbalance for all the informative loci. Fluorescence in situ hybridization and Southern analysis were performed to test some candidate oncogenes in 8q24 (MYC) and 15q25-qter (IGF1R and FES). Two of seven tumors showed high-level amplification of MYC relative to the centromere (> 3-fold), another two tumors had low-level amplification (1.5- to 3.0-fold), and one displayed 5.5 MYC signals/cell. In relation to the FES gene, low-level amplification was found in three tumors. Southern analysis showed five cases with a low-level amplification of IGF1R. Our data suggest that either few extra gene copies may be enough for cancer progression or other genes located in these regions are responsible for the amplifications found by CGH.     

Overrepresentation of the short arm of chromosome 12 in seminoma and nonseminoma groups of testicular germ cell tumors

Histopathological differentiation between dermatofibrosarcoma protuberans (DFSP) and dermatofibroma (DF) is often difficult, because both neoplasms share some clinical features and the presence of a storiform pattern. In the present study, we investigated the usefulness of comparative genomic hybridization (CGH) in the diagnosis of these entities by examining 12 DFSP and 12 DF cases. The most frequent DNA sequence copy number changes detected in 10 (83%) of 12 DFSP cases (mean, 1.9 aberrations/tumor; range, 0-3) consisted of gains of 17q22-qter (10 tumors), 22q13 (nine tumors), and 8q24.1-qter (three tumors). High-level amplification, which was detected in three tumors, was seen only in chromosome 17, with 17q23-q25 as the minimal common region. Loss of DNA sequences was not found in DFSP cases. In contrast, two (17%) of the 12 DF cases (mean, 0.5 aberrations/tumor; range, 0-4) showed DNA sequence copy number changes, although recurrent gains and losses and high-level amplifications were not observed. Gains were more common than losses in DF. Overrepresentation of 17q and 22q sequences was a common finding in DFSP but not in DF. Thus, CGH seems to be useful for distinguishing DFSP from DF in most cases.     

Cytogenetic findings in two synovial sarcomas

Cytogenetic analysis was performed after short-term tissue culture of two recurrent synovial sarcomas. The tumors were classified on the basis of morphology, location, and immunohistochemistry. In a poorly differentiated tumor, the karyotype 49,XY, +7, +8, +19,t(5:18) (q11.2;q11.2), and in a biphasic tumor two clonal cell lines with common translocations t(X;18)(p11.2;q11.2) and t(12;17)(p11.2;q11.2) were present. In the predominant cell line several other structural aberrations including t(1;12)(q21;q24.3), t(3;18)(p23;q21), and 17p+ were found. A comparison of our results with previously published studies suggests that in addition to t(X;18), translocations of chromosome 18 with other chromosomes may represent a consistent feature of chromosomal changes in synovial sarcoma.     

Immunohistochemical staining for TLE1 distinguishes synovial sarcoma from histologic mimics

Transducer-like enhancer of split 1 (TLE1) is overexpressed in synovial sarcomas. We investigated TLE1 expression by immunohistochemical analysis in a well-characterized series of synovial sarcomas and other mesenchymal tumors most commonly considered in the differential diagnosis. Whole tissue sections of 212 tumors were evaluated: 73 synovial sarcomas (23 biphasic, 28 monophasic, 22 poorly differentiated), 47 malignant peripheral nerve sheath tumors (MPNSTs), 49 solitary fibrous tumors (SFTs), 20 fibrosarcomatous variants of dermatofibrosarcoma protuberans, and 23 Ewing sarcomas/primitive neuroectodermal tumors (PNETs). All monophasic and poorly differentiated SSs and Ewing sarcoma/PNETs were previously confirmed to harbor t(X;18) and EWSR1 gene rearrangements, respectively. In total, 60 (82%) of 73 synovial sarcomas were positive for TLE1, including 18 biphasic (78%), 22 monophasic (79%), and 20 poorly differentiated (91%) tumors. Of the other tumors, only 7 MPNSTs (15%) and 4 SFTs (8%) were positive for TLE1, most of which showed only weak staining. TLE1 is a sensitive and specific marker for synovial sarcoma and can be helpful to distinguish synovial sarcoma from histologic mimics, particularly if moderate or strong staining is observed. In this study, only a small subset of MPNSTs and SFTs showed limited staining for TLE1.     

Interphase cytogenetic analysis of distinct X-chromosomal translocation breakpoints in synovial sarcoma

Synovial sarcomas show a specific translocation involving chromosomes X and 18, t(X;18)(p11.2;q11.2). Two distinct X-chromosomal breakpoints occur in different synovial sarcoma tumour samples. These breakpoints are located within two related genomic regions containing ornithine aminotransferase-like sequences, termed OATAL1 and OATL2. Preliminary observations indicated the potential correlation of OATL1-associated breakpoints with biphasic tumours and OATL2-associated breakpoints with monophasic fibrous tumours. The present study uses interphase cytogenetics to investigate the nature of chromosomal aberrations in frozen synovial sarcoma tissue samples. Two-colour fluorescence in situ hybridization (FISH) was performed using probes specific for the centromeres of chromosome X or 18, along with yeast artificial chromosome probes corresponding to the distinct breakpoint regions on Xp. One monophasic epithelial and two monophasic fibrous synovial sarcomas showed an OATL2-associated breakpoint, while a biphasic tumour revealed a hybridization pattern indicating a breakpoint within the OATL1 region. These results confirm our previous suggestion of a relationship between alternative breakpoints in Xp11.2 and different histological phenotypes observed in synovial sarcomas. They also demonstrate the utility of the two-colour hybridization approach for the identification of chromosomal changes in interphase nuclei isolated from frozen tissues.     

Synovial sarcoma. Inter-relationship of the biphasic and monophasic subtypes.

In order to assess minimum diagnostic criteria for synovial sarcoma, particularly the monophasic variety, and the inter-relationship between the monophasic and biphasic types, 32 examples were studied histologically, immunohistochemically (26 cases), and ultrastructurally (13 cases). Of the six biphasic synovial sarcomas examined by electron microscopy, the spindle cell component did not show evidence of epithelial differentiation or resemble the epithelial phase, but did appear fibroblastic; no tumor cells transitional between the spindle and epithelial component were evident. In contrast, all of the seven monophasic lesions had ultrastructural growth patterns and some cellular features approximating the epithelial cells of the biphasic variant. In 11 biphasic synovial sarcomas, epithelial membrane antigen was detected in the glandular epithelium of all cases and cytokeratins in eight cases; in no case were these antigens detected in the spindle cell regions of biphasic lesions. Of the 15 monophasic synovial sarcomas, two were positive for cytokeratins and four for epithelial membrane antigen. Thus, the detection of epithelial markers either immunohistochemically or by electron microscopy (or both) should be the minimal diagnostic criteria for monophasic synovial sarcomas. Based on these findings, it is suggested that monophasic synovial sarcomas do not represent the spindle cell or "stromal" phase of biphasic synovial sarcomas, but are a poorly differentiated variant of the latter. As others have suggested, these tumors are, in fact, carcinosarcomas and carcinomas of the soft tissues and the designation synovial sarcoma is inappropriate for this tumor class.     

Gains of 12q are the most frequent genomic imbalances in adult fibrosarcoma and are correlated with a poor outcome

Comparative genomic hybridization was used to analyze 41 adult fibrosarcomas from 34 patients. Thirty-one patients showed in their tumors DNA sequence copy number changes (mean 11, range 3-25). The minimal common regions for the most frequent gains were narrowed down to 12q21 (18 cases); 12q14-q15 and 14q22 (16 cases each); 4q22, 7q31, and 14q23-q24 (15 cases each); and 4q21, 4q23-q24, 8q22, and 12q22 (14 cases each). Twenty-five high-level amplifications were observed in 12 samples. 12q21 and 18p were affected three times each; and 1p21, 4q31.3, 7p21, 12q14-q15, Xp22.1-p22.2, and Xq22-q23 two times each. Losses were less frequent than gains. Early stages of adult fibrosarcomas were characterized by frequent gains of chromosomes 2, 4q, and 14q, whereas gains of chromosomes 7 and 8q were associated with progression. Gains of 12q were frequent in all of the developmental steps of this soft-tissue sarcoma. By investigation of several tumors of the same patient, a number of corresponding changes were always detected. Adult fibrosarcomas from patients who died during the observation time showed statistically significant more frequent gains of 8q, 12q, 13q, and 15q compared to the fibrosarcomas of patients who are alive. Gains and high-level amplifications of 12q14-q22, which were the most frequent genomic imbalances, partly reflected an MDM2 amplification, indicating the importance of this region in the tumorigenesis of sarcomas. In adult fibrosarcomas, a gain of 12q22 correlated significantly (P = 0.028) with a poor overall survival rate.     

Comparative genomic hybridization analysis reveals 3q gain resulting in genetic alteration in 3q in advanced oral squamous cell carcinoma.

We analyzed DNA sequence copy number aberrations (DSCNAs) in 17 primary oral squamous cell carcinomas (OSCCs) by comparative genomic hybridization. DSCNAs were detected frequently at 3q25-qter (7/17), Xp21 (5/17), and Xq12-q23 and 8q23-q24 (4/17), and losses were detected frequently at 13q21-q22 (5/17), 3p21-pter, 4p15-pter and 17p13 (4/17), and 8p22-pter and 9p21-pter (3/17). Four tumors showed amplifications of seven loci: 3q11-qter, 3q13, 3q26, 7q21-q22, 8q23-qter, 9p22-pter, and 12p11. The total number of DSCNAs was significantly greater in stage III and stage IV tumors than in stage I and stage II tumors (P=.008). Furthermore, 3q gain was detected preferentially in stage III and stage IV tumors (6/8) rather than in stage I and stage II tumors (1/9, P=.013). In our study, all tumors with gain of 3q also contained one or more loss(es) in common regions. On the other hand, all tumors with gain of 9p did not contain 3q gains. These observations indicate that gain of 3q and accumulation of DSCNAs are strongly associated with tumor progression in OSCC. Furthermore, 3q gain and loss of one or more additional loci in common aberration regions appears to be a group of DSCNs associated with dominant genetic pathways of leading to advanced OSCCs.     

Comparative cytogenetic study of spindle cell and pleomorphic leiomyosarcomas of soft tissues: a report from the CHAMP Study Group

Leiomyosarcomas (LMS) of soft tissues frequently show complex karyotypic changes, and no specific aberration has been identified. The aim of this study was to search for recurrent chromosome aberrations in soft tissue LMSs and to correlate these, if present, with morphological and clinical parameters. From a series of soft tissue sarcomas thoroughly reexamined cytogenetically and histopathologically, 45 LMSs were retrieved; 35 were classified microscopically as spindle cell, 3 as epithelioid, and 7 as pleomorphic. Clonal chromosome changes were present in 14, 3, and 3 cases, respectively. This series was combined with 11 previously published, karyotypically abnormal pleomorphic LMSs for cytogenetic-clinico-histopathological correlations. The breakpoints were widely scattered, with no predilection of any of the recurrent breakpoints and losses to any of the morphologic subtypes. Combining numerical and unbalanced structural changes, the most frequently lost segments were 3p21-p23 (11 cases), 8p21-pter, 13q12-q13, 13q32-qter (10 cases each), 1q42-qter, 2p15-pter, 18p11 (9 cases each), 1p36, 11q23-qter (8 cases each), and 10q23-qter (7 cases). The most frequent gain was 1q12-q31 (6 cases). There was a greater frequency of losses in 1p and 8p and a lower frequency of losses in 10q and 13q in tumors that had metastasized than in localized tumors. We conclude that LMSs with clonal abnormalities display highly complex karyotypic changes and extensive heterogeneity. No significant correlation exists between these changes and age and sex of the patients, or with depth of tumor, topography, microscopic subtype, or tumor grade. Losses in 1p36 and 8p21-pter may be associated with increased risk of metastases. Comparison of our findings in soft tissue LMS with those previously reported in LMS in other locations suggest that the karyotypic profile is more dependent on site of origin than on microscopic features.     

Analysis of Genomic Alterations in Sporadic Adrenocortical Lesions : Gain of Chromosome 17 Is an Early Event in Adrenocortical Tumorigenesis

Genetic changes underlying the tumorigenesis of sporadic adrenocortical tumors are poorly characterized. To search for characteristic genomic imbalances involved in adrenocortical tumors, we examined 41 adrenocortical lesions (12 carcinomas, 23 adenomas, and 6 hyperplasias) by comparative genomic hybridization. Our results show that genetic alterations are more frequent in malignant than in benign lesions and that they rarely occur in hyperplastic lesions. The most frequent DNA copy number changes in adrenocortical carcinomas included losses of 1p21-31, 2q, 3p, 3q, 6q, 9p, and 11q14-qter, as well as gains and amplifications of 5q12, 9q32-qter, 12q, and 20q. The genomic aberrations prevalently occurring in adrenocortical adenomas were gains of 17q, 17p, and 9q32-qter. Gains found in 2 of 6 adrenocortical hyperplastic lesions involved chromosome 17 or 17q only. These data indicate that oncogenes determining the early tumorigenesis of adrenocortical tumors may exist on chromosome 17 and that the number of genomic alterations is closely associated with tumor behavior in adrenocortical tumors.     

Gains, losses, and amplifications of DNA sequences evaluated by comparative genomic hybridization in chondrosarcomas.

Comparative genomic hybridization was used to search for previously unknown gains and losses of DNA sequences along all chromosome arms in 29 chondrosarcoma specimens obtained from 23 patients. Extensive genetic aberrations, with a mean of 6 changes per tumor (range, 1 to 24), were detected in 21 of the 29 samples analyzed (72%). The majority of these changes were gains of whole chromosomes or whole chromosome arms. Gains of DNA sequence copy number were most frequent at 20q (38%), 17p (38%), 20p (31%), 1cen-q24 (28%), and 14q23-qter (28%). High-level amplifications of small chromosome regions were sporadic, detected in only 17% of the samples. The only recurrent high-level amplification, seen in two tumors (7%), affected the minimal common region 12cen-q15. Other amplifications, each encountered only once, involved 1p33-p35, 2p23-pter, 4p, 6p22-pter, 18q12-q22, 19p13.2, 19q13.2, and 20q13.1. Losses of DNA sequences were rare and were most commonly observed at 6cen-q22 (17%) and 9p (17%).     

Gains and losses of DNA sequences in liposarcomas evaluated by comparative genomic hybridization

Comparative genomic hybridization (CGH) was used to detect and map the regions of gain, high-level amplification, and loss of DNA sequences in 14 liposarcomas. Thirteen tumors showed DNA sequence copy number changes of one or more genomic regions (mean, six aberrations/tumor; range, 0–17). These aberrations were observed in almost every chromosome but some chromosomal regions were affected more often than others. DNA sequence gains were more frequent than losses. The most common gain was seen at 12q14-21 (50% of tumors). Other frequent gains (29%) were of Iq21-24, 8cen-q21.2, 19q, and 20q. High-level amplification was observed in six (43%) tumors and included as minimal common segments bands 12q15, Iq22, and Iq24. In five (36%) tumors, sequences at Iq21-24 and Iq32 were found to be gained simultaneously with 12q14-21, which means that in 71% of the tumors with gain at 12q, an increase of DNA sequence copy number at Iq was also observed. The most common losses of DNA sequences (21%) occurred from regions 9p21-pter and 13q21-qter. Most of the aforementioned regions have not previously been reported to be altered in liposarcomas. The detection of a novel recurring amplicon at Iq21-24 with high-level amplification at Iq22 and frequent simultaneous DNA sequence gain at 12q14-21 (high-level amplification at 12q15) suggests that genes linked to both these regions may play a significant role in the development and progression of liposarcomas. Genes Chromosom Cancer 15:89–94 (1996). © 1996 Wiley-Liss, Inc.     

Chromosomes in the diagnosis of soft tissue tumors. I. Synovial sarcoma.

It has been established that nonrandom chromosome rearrangements are characteristic of specific types of neoplasia. We present six new cases of sarcoma that had in common the same chromosome abnormality, i.e., a balanced translocation between chromosomes X and 18, t(X;18)(p11.2;q11.2), and evaluate the 15 cases with this translocation in the literature. The histological diagnosis was synovial sarcoma in 19 cases and malignant fibrous histiocytoma and fibrosarcoma in the remaining two tumors, respectively. The translocation was found in tumors of both the biphasic and monophasic types, as well as in poorly differentiated synovial sarcoma. The two nonsynovial sarcomas with the t(X;18) were described as spindle cell tumors but failed to show the presence of cytokeratins by immunohistochemical stains. Even with the numerous variabilities on which this test depends, the cytogenetic analysis holds great promise as a tool for the diagnosis of synovial sarcoma.     

Comparative genomic hybridization reveals complex genetic changes in primary breast cancer tumors and their cell lines

DNA copy number changes were characterized by comparative genomic hybridization (CGH) in 18 breast cancer cell lines. In 5 of these, the results were comparable with those from the primary tumors of which the cell lines were established. All of the cell lines showed extensive DNA copy number changes, with a mean of 16.3 +/- 1.1 aberrations per sample (range 7-26). All of the cell lines had a gain at 8q22-qter. Other common gains of DNA sequences occurred at 1q31-32 (89%), 20q12-q13.2 (83%), 8q13 (72%), 3q26.1-qter (67%), 17q21-qter (67%) 5p14 (61%), 6p22 (56%), and 22pter-qter (50%). High-level amplifications were observed in all cell lines; the most frequent minimal common regions were 8q24.1 (89%), 20q12 (61%), 1q41 (39%), and 20p11.2 (28%). Losses were observed less frequently than gains and the minimal common regions of the most frequent losses were Xq11-q12 (56%), Xp11.2-pter (50%), 13q21 (50%), 8p12-pter (44%), 4p13-p14 (39%), 6q15-q22 (39%), and 18q11.2-qter (33%). Although the cell lines showed more DNA copy number changes than the primary tumors, all aberrations, except one found in a primary tumor, were always present in the corresponding cell line. High-level amplifications found both in primary tumors and cell lines were at 1q, 8q, 17q, and 20q. The DNA copy number changes detected in these cell lines can be valuable in investigation of tumor progression in vitro and for a more detailed mapping and isolation of genes implicated in breast cancer.     

Patterns of keratin polypeptides in 110 biphasic, monophasic, and poorly differentiated synovial sarcomas

Synovial sarcoma is a mesenchymal neoplasm of unknown histogenesis that shows various degrees of epithelial differentiation. It is known to contain simple epithelial keratins, and the possibility of complex epithelial keratin expression has been suggested. In this study, we immunohistochemically examined 110 well-documented synovial sarcomas including 44 biphasic, 48 monophasic, and 18 poorly differentiated (undifferentiated, highly mitotically active) tumors for 11 different keratin (K) polypeptides of the Moll catalogue. The epithelia of biphasic synovial sarcomas showed consistent, extensive reactivity for K7, K8, K14, K18, and K19. Other keratins seen in the epithelia of biphasic tumors included K17 (variable, in 77%), K13 (25%), K16 (23%), and K6 (24%) in the minority of biphasic tumors, predominantly in stratified-appearing epithelia. K10 was detected only focally in one case that showed keratinizing squamous differentiation. Focal expression of K20 was seen in 27% of cases. Monophasic synovial sarcomas had a more limited keratin repertory. Simple epithelial keratin positivity was detected, usually focally for K7 (79%), K19 (60%), K8 (45%), and K18 (46%). Two cases showed more extensive keratin positivity in the spindle cells. The monophasic tumors showed limited positivity for complex epithelial keratins: K14 (28%) and K17 (10%). K20 was detected focally in 6% of the monophasic tumors; other keratins were not detected. The poorly differentiated synovial sarcomas showed limited simple epithelial keratin reactivity, usually limited to scattered cells: K19 (61%), K7 (50%), K18 (47%), K8 (33%), but five cases showed more extensive positivity. Complex epithelial keratins were scant: K14 in one case and K17 in two cases. The immunoreactivity of capillary endothelia seen for K7 and K18 (but not for K8 and K19 with the antibodies used) is a potential diagnostic pitfall, and may cause overdiagnosis of synovial sarcoma if not properly recognized. In summary, we show complex patterns of keratins in synovial sarcoma, especially in the biphasic tumors. Such patterns establish a baseline in differential diagnostic considerations, and give an insight into the complex epithelial differentiation of this enigmatic mesenchymal tumor. Received: 17 February 2000 / Accepted: 3 April 2000