Virchows Archiv

, Volume 452, Issue 3, pp 259–268

Molecular genetic aberrations of ovarian and uterine carcinosarcomas—a CGH and FISH study

  • Alexander Schipf
  • Doris Mayr
  • Thomas Kirchner
  • Joachim Diebold
Original Article

DOI: 10.1007/s00428-007-0557-6

Cite this article as:
Schipf, A., Mayr, D., Kirchner, T. et al. Virchows Arch (2008) 452: 259. doi:10.1007/s00428-007-0557-6

Abstract

The origin of carcinosarcomas of the ovary and uterus has long been discussed. In this study, we used a molecular–genetic approach to elucidate the tumorigenesis of carcinosarcomas of these organs correlating our findings with the specific biphasic pattern of these tumors. We analyzed a series of 30 paraffin-embedded carcinosarcomas of the ovary and the uterus using comparative genomic hybridization (CGH) and fluorescence in-situ hybridization (FISH). In general, gains (85%) were observed more frequently, than losses (30%). Characteristic and frequent chromosomal amplification was observed on chromosome 8q and 20q (42 and 70%). FISH revealed c-myc (8q24.12) and ZNF217 (20q13.2) amplification in 78 and 87%. Amplification of ZNF217 was mostly seen in both tumor components, whereas amplification of c-myc was observed less often in the sarcomatous than in the carcinomatous tumor component. Analysis of the proliferation index using Ki67 immunohistochemistry revealed a strong or moderate expression in all cases, wherein the carcinomatous tumor component showed significantly a higher proliferation index compared to the sarcomatous tumor areas. Although our results are in agreement with a monoclonal origin of ovarian and uterine carcinosarcomas, the carcinomatous component seems to be the more aggressive part of the tumor. Furthermore, the observed patterns of genetic aberrations are highly similar to those of serous carcinomas. This is compatible with the current opinion that these neoplasms should be considered as metaplastic carcinomas.

Keywords

Carcinosarcoma CGH FISH Immunohistology Ovary 

Introduction

Carcinosarcomas of the female genital tract, also called malignant mixed Mullerian tumors, are highly aggressive biphasic neoplasms composed of carcinomatous and sarcomatous components. They usually occur in the uterus, ovaries, Fallopian tubes, and vagina, in descending order of frequency. Histologically, the carcinomatous component is usually of the endometrioid, serous, or clear-cell type. The sarcomatous component may be either homologous (endometrial stromal sarcoma, fibrosarcoma, leiomyosarcoma) or heterologous. These uncommon tumors are usually found in postmenopausal women, frequently at an advanced stage, and have a very poor prognosis.

The histogenesis of carcinosarcomas of the female genital tract has been an issue of great controversy. To explain the characteristic biphasic pattern, several theories have been proposed with two major theories being favored. One theory suggests that these tumors develop through two independent neoplastic processes, forming a biclonal tumor (“collision” theory). According to the other theory, the tumor consists of a monoclonal tumor cell population undergoing partially metaplastic transformation (“conversion” theory). Other histogenetic theories, such as the “combination” theory where both tumor components derive from a single stem cell, which differentiates very early in tumorigenesis in diverging directions, or the “composition” theory, interpreting the sarcomatous component as pseudosarcomatous stromal reaction, were largely dismissed [27].

In the last few years, evidence in favor of a monoclonal origin particularly supporting the conversion hypothesis has accumulated, based on a number of studies using sensitive morphological, cytogenetic, ultrastructural, and immunohistochemical techniques [2, 4, 6, 8, 13, 16, 29, 30].

Cytogenetic studies of carcinosarcomas have revealed extremely complex karyotypes with multiple chromosomal gains and losses. Multiple chromosomal alterations were described, such as polysomy 8, loss of heterozygosity (LOH), or p53 mutations. Deoxyribonucleic acid (DNA) cytometry showed nondiploidy in most carcinosarcomas [1, 3, 7, 21, 25, 26, 27, 28, 35].

All these studies were based on small number of cases or in vitro cell cultures. The genetic mechanism determining the intratumoral heterogeneity of the biphasic carcino-sarcomatous phenotype is still unknown.

A comprehensive analysis of carcinosarcomas by comparative genomic hybridization (CGH) and fluorescence in situ hybridization (FISH) has not been performed so far. Therefore, in the present study, we investigated the molecular aberrations in 30 carcinosarcomas by these techniques. We correlated our findings with the biphasic phenotype to elucidate the clonal relationship between the epithelial and mesenchymal tumor component and the possible mode of development of the biphasic pattern. Our observations provide evidence that the two tumor components are very closely related and support the “conversion” theory.

Materials and methods

Specimen

Formalin-fixed and paraffin-embedded tissue of 30 patients diagnosed with carcinosarcoma of the ovary and uterus were retrieved from archives of the Institute of Pathology of the University of Munich.

Twenty-four cases were located in the ovary, six cases in the endometrium. Ten tumors showed heterologous differentiation of the sarcomatous component. All the other carcinosarcomas were of homologous differentiation.

All patients had been treated surgically between 1988 and 2004 at the same institution (Department of Gynecology of the Klinikum Grosshadern, University of Munich). No patient received any other treatment before the operation.

According to the criteria of the International Federation of Gynecologists and Obstetricians and the International Union against Cancer, they were stage IC to IVB.

Per case, between 1 and 20 representative histological slides were available.

Comparative genomic hybridization

Chromosome preparation and karyotype analysis were performed as previously described [34]. Total tumor DNA was extracted from tissue sections—without differentiation of carcinomatous or sarcomatous components. One microgram of tumor DNA was labeled by a standard nick-translation reaction with biotin-16-deoxyuridine triphosphate (dUTP) and 1 μg reference DNA from a healthy male donor with digoxigenic-11-dUTP (each from Boehringer Mannheim). The purification of the labeled DNA fragments of less than 100 bp was achieved by applying column chromatography (Sephadex-G50, Amersham Biosciences Pharmacia GE Healthcare, Munich). Repetitive sequences were blocked with 70 μg of Cot 1 DNA.

Gray-band images of 4′,6-diamidino-2-phenylindole, fluorescein isothiocyanate, and rhodamine fluorescence were taken using a charge-coupled-device camera coupled to a Zeiss microscope. For digital image analysis and subsequent karyotyping, the ISIS QUIPS-XL (MetaSystems, Altlussheim) software package was used. Only metaphase spreads showing a high green-to-red intensity were taken in account. Between 8 and 20 metaphases (mean = 12.2) for each chromosome were analyzed. Corresponding ratio profiles were evaluated only within the 99% confidence limit. The 50% threshold (upper threshold, 1.2; lower threshold, 0.8) were applied to define the chromosomal regions of DNA sequence losses and gains.

Fluorescence in situ hybridization

After removal of excess normal tissue, FISH was performed on tumor tissue sections. To recognize and distinguish carcinomatous or sarcomatous tumor components, FISH sections were carefully compared with serial hematoxylin–eosin-stained sections.

The in situ hybridization was performed as described previously [10]. The hybridization mix was prepared according to the manufacturer’s protocol with DNA-specific probes for 8q24.12 and 20q13.2 (c-myc and ZNF217, Abbott/Vysis, Wiesbaden) [5].

The slides were evaluated under a ZEISS (Göttingen, Germany) axioscope fluorescence microscope equipped with an HBO 100-W mercury lamp and dual- and triple-band pass filters (Applied Imaging, Vysis, AF Analysentechnik, Tuebingen). Hybridization signals were counted in 60 to 120 nuclei per case. Only discrete signals in nonoverlapping nuclei with distinct nuclear borders were evaluated. Split signals were counted as one signal. Amplification was defined as a ratio of gene to centromere signals of greater than 2, whereas polysomy was defined as an increase of gene copy numbers with a constant gene-to-centromere ratio less than 2 in all examined tumor cells. In 13 cases, the enumeration of FISH signals was performed twice. The difference between the first and the second count was 0.25 ± 0.21 hybridization signal per nucleus for centromere signals and 0.33 ± 0.35 for gene signals. Thus, the results proved to be highly reproducible.

Immunhistochemistry

All paraffin-embedded specimens were cut at 2–3 μm and mounted on SuperFrost/Plus microscope slides (Menzel, Germany). After deparaffinization and rehydration, sections were immunhistochemically stained with a streptavidin–biotin complex system (Dako, Hamburg). Anticytokeratin monoclonal antibodies (Dako) were used for detection of the epithelial component. To detect the mesenchymal tumor components, antivimentin monoclonal antibody (Dako) and antidesmin monoclonal antibody (Dako) were used.

For Ki67, the monoclonal antibody MIB1 (Dako, Glostrup, Denmark) was used at a dilution of 1:50. The immunohistochemical results of Ki67 were scored quantitatively. Carcinomatous and sarcomatous tumor cells were counted, and the percentage of the immunoreactive cells was determined.

Results

Histology

Histologically, all tumors were composed of an admixture of malignant epithelial and mesenchymal stromal cells. To differentiate into distinct tumor components, immunohistochemistry was performed in all cases (Fig. 1). Epithelial and mesenchymal neoplastic cells were clearly distinguished.
Fig. 1

Representative hematoxylin–eosin-stained and immunhistochemistry sections of a carcinosarcoma. Epithelial components were marked by keratin staining, mesenchymal by vimentin

Histological features of investigated carcinosarcomas are summarized in Table 1. Sixteen of 24 ovarian and four of six uterine tumors showed homologous differentiation of the sarcomatous component such as endometrial stromal sarcoma, fibrosarcoma, and leiomyosarcoma. The carcinomatous component was of endometrioid or serous type.
Table 1

Stage, age at time of diagnosis, tumor localization, and histological differentiation of the carcinomatous and sarcomatous components (n = 30)

Case number

Stage

Age

Localization

Epithelial differentiation

Sarcomatous differentiation

1

IIIB

70

Ovarian

Serous

Homologous

2

IIIC

75

Ovarian

Endometrioid

Heterologous

3

IIB

59

Ovarian

Serous

Homologous

4

IC

64

Ovarian

Serous

Homologous

5

IIIC

80

Uterine

Serous

Homologous

6

IIB

77

Ovarian

Serous

Homologous

7

IIA

73

Ovarian

Serous

Heterologous

8

IIIC

68

Uterine

Serous

Homologous

9

IIA

68

Uterine

Serous

Homologous

10

IIIB

42

Ovarian

Endometrioid

Heterologous

11

IIIB

73

Ovarian

Serous

Homologous

12

IVB

48

Ovarian

Serous

Homologous

13

IIIC

67

Ovarian

Endometrioid

Homologous

14

IIIC

68

Uterine

Serous

Homologous

15

IIIC

63

Ovarian

Endometrioid

Homologous

16

IIIC

79

Ovarian

Serous

Homologous

17

IVB

81

Ovarian

Serous

Heterologous

18

IIIB

74

Ovarian

Endometrioid

Heterologous

19

IIIC

87

Uterine

Serous

Heterologous

20

IVB

72

Ovarian

Endometrioid

Homologous

21

IIICB

79

Uterine

Serous

Heterologous

22

IVB

X

Ovarian

Serous

Heterologous

23

IVB

48

Ovarian

Serous

Homologous

24

IIIB

67

Ovarian

Serous

Homologous

25

IIIB

70

Ovarian

Serous

Homologous

26

IIIB

65

Ovarian

Endometrioid

Homologous

27

IIIB

53

Ovarian

Serous

heterologous

28

IIIC

75

Ovarian

Serous

heterologous

29

IVB

79

Ovarian

Endometrioid

Homologous

30

IVA

68

Ovarian

Serous

Homologous

In seven cases, the carcinomatous tumor component predominated, whereas in 18 cases, sarcomatous areas prevailed.

Comparative genomic hybridization

To get a general idea of chromosomal aberrations in the carcinomatous and sarcomatous tumor components, we analyzed 25 cases using CGH technology. In general, the tumors showed a large number of chromosomal imbalances, and almost all chromosomes were affected (Fig. 2).
Fig. 2

Ideogram showing comparative genomic hybridization findings in 25 carcinosarcomas of the ovary and uterus (red bars on the left of chromosomes are losses, and green bars on the right are gains)

Gains (85%) were observed more frequently than losses (30%). The most frequently occurring CGH changes were gains on chromosomes 1q, 2p, 8q, 12p, 19q, and 20q. Wherein, amplifications on chromosome 8q and 20q were particularly frequent (42 and 70%) and were correlated with each other in more than 50%. Losses were observed most frequently on chromosome 4q, 9q, and 13q. CGH profiles of cases with predominance of carcinomatous areas were highly similar to CGH ideograms of cases with preponderance of sarcomatous tissue (Fig. 3).
Fig. 3

Two examples of CGH ideograms of a tumor with predominance of carcinomatous and sarcomatous areas, respectively, revealing a high grade of similarity

Fluorescence in situ hybridization

FISH analysis for chromosome 8 and 20 was successful in 23 of 30 cases. Many carcinosarcomas contained aberrations on chromosomes 8 and 20 detectable by FISH.

Increase of c-myc gene copy number and a gene-to-centromere ratio less than 2 in all examined tumor cells (polysomy) was observed in five tumors. Amplification of c-myc (gene/centromere signals greater than 2) was observed in 18 cases (Table 2A).
Table 2

Overview of c-myc and znf217 FISH analysis (n = 23, polysomy: increase of gene copy number and constant gene to centromere ratio less than 2; amplification: gene to centromere index greater than 2, A), comparison of genetic aberrations in carcinomatous and sarcomatous components of tumors with gene amplification (B)

 

c-myc

znf217

A

tumors with polysomy

5

4

tumors with amplification

18

19

Total

23

23

B

Cases with gene amplification in

  

 Carcinomatous area only

14

14

 Sarcomatous area only

4

15

 Carcinomatous and sarcomatous area

2

10

Total

18

19

Comparing gene and centromere 8 copy numbers, 16 of the 18 cases showed amplification for c-myc in the carcinomatous tumor component. The median number of c-myc hybridization signals per nucleus was three in five cases and greater than three in nine cases. In contrast, c-myc amplifications in the sarcomatous component were observed in only four (of 18) cases wherein two cases showed amplification in both tumor components simultaneously (Table 2B, Figs. 4 and 5).
Fig. 4

Fluorescence in situ hybridization analyses. Examples for 20q13 and 8q24 in sarcomatous and carcinomatous tumor components. Three to six nuclei display multiple, red hybridization signals in both components for 20q13 and 8q24 hybridization signals amplified in the carcinomatous tumor component

Fig. 5

Comparison of c-myc and znf217 gene copy numbers in carcinomatous and sarcomatous areas of tumors with gene amplification

For znf217 analysis, the median number of hybridization signals per nucleus was three and higher in 19 cases. Four tumors showed polysomy, and in 19 cases, we observed amplification of znf217 chromosome regions. In contrast to c-myc amplifications, a high level of homology was seen between the two components in each tumor. Fourteen of 19 cases showed amplification of znf217 in the carcinomatous and 15 cases in the sarcomatous tumor area, whereas simultaneous amplification in both tumor components was seen in ten (of 19) cases (Table 2B, Fig. 5).

Comparison of CGH and FISH

Comparison of CGH and FISH results showed that most of the chromosomal gains in CGH on chromosomes 8 and 20 could be detected by FISH (52 and 70%). Negative findings in CGH analysis also correlated with negative FISH results. Gains on chromosome 8q24 (c-myc) were observed on CGH analysis in 14 cases, and nine tumors showed no amplification in CGH. Using FISH analysis, we detected among these negative tumors six cases with amplification on chromosome 8q24. For znf217 (20q13), the majority of gene amplifications could be demonstrated in both techniques (Table 3). The Chi-squared test revealed a statistically significant correlation (p < 0.05).
Table 3

Comparison of CGH and FISH results

 

Positive

Negative

FISH 8q24 (n = 23)

CGH 8q24

Positive

12

2

Negative

6

3

 

Fish 20q13 (n = 23)

CGH 20q13

Positive

16

1

Negative

4

2

CGH amplifications for 8q24.12 and 20q13.2 (rows) and FISH gene/centromere index (columns), n = 23

Chi-squared test, p < 0.05

Immunohistochemistry

Twenty-five cases were analyzed in epithelial and mesenchymal areas for the proliferation-associated antigen Ki67 to assess the growth fraction in both tumor areas. Ki67 was strongly or moderately expressed in all cases, wherein the carcinomatous tumor component showed a significantly (p < 0.001) higher proliferation index (mean 70%) compared to the sarcomatous tumor areas (mean 28%; Table 4, Fig. 6).
Fig. 6

Representative immunohistochemical detection of nuclear Ki67 accumulation in carcinomatous (a) and sarcomatous (b) tumor component. The original magnification of each, ×40 objective

Table 4

Proliferation index in carcinomatous and sarcomatous tumor components (immunhistochemical Ki67 staining)

Case number

Carcinomatous tumor area (Ki67 staining, %)

Sarcomatous tumor area (Ki67 staining, %)

1

80

5

3

75

20

4

75

40

5

60

33

6

80

10

7

80

20

8

50

20

9

45

10

11

60

25

12

60

10

13

40

40

14

90

50

15

90

20

16

80

40

17

60

40

20

75

10

21

75

70

22

60

50

23

80

20

24

60

33

25

80

10

27

60

50

28

75

25

29

80

20

30

80

40

Discussion

Despite many histological, immunohistochemical, cytogenetic, and ultrastructural analyses, the histogenesis of carcinosarcomas still remained a matter of controversy [2, 4, 16, 17, 18, 19, 20, 22, 23, 29, 33, 36].

In this study, we used a molecular–genetic approach to elucidate the tumorigenesis of carcinosarcomas of the ovary and the uterus correlating our findings with the biphasic pattern of these tumors. For the first time, a combination of CGH and FISH techniques was used for this purpose.

Histologically, the tumors included in this study showed a great variety of epithelial and mesenchymal differentiation. Most tumors were of advanced stage, and the majority showed homologous differentiation of the sarcomatous tumor component.

Investigating the cytogenetic pattern of the carcinosarcomas, we used molecular–genetic techniques to characterize chromosomal changes with special consideration of the intratumoral heterogeneity of carcinosarcomas. It is well known that chromosomal aberrations occur commonly in carcinosarcomas of the female genital tract. Cytogenetic studies have shown that p53 and K-ras mutations, LOH, and microsatellite instability arise in both tumor components [1, 14, 15, 25, 38]. Wada et al. [38] found identical mutations of p53 and K-ras in the two different tumor components. Identical p53 point mutations were also observed in both tumor areas, by other groups supporting a common monoclonal origin of both carcinoma and sarcoma [25, 37].

p53 mutations seem to be an early event in tumorigenesis occurring before the separate development of epithelial and mesenchymal differentiation. p53, C-erbB-2, and c-myc protein expression has been shown in the carcinomatous as well in the sarcomatous tumor areas [8, 9, 24, 25, 26].

Regarding LOH, molecular analysis frequently showed aberrations, which were almost identically in the two tumor components [1].

Emoto et al. [12] reported the results of cell culture studies and analysis of the c-myc gene revealing concordant karyotype abnormalities and c-myc amplifications in the carcinomatous and sarcomatous subclones.

On cytogenetic level, our results demonstrate a uniform pattern of chromosomal gains and losses on the majority of the chromosomes in CGH analysis. Most aberrations were clustered on distinct chromosomal segments, and gains (chromosome 1q, 2p, 8q, 12p, 19q, and 20q) were observed more frequently than losses (4q, 9q, and 13q). The amplicons mentioned contain a number of well known oncogenes, e.g., BTAK (20q), KRAS (12p), or PIK3CA (3q).

Because total tumor DNA was used, carcinoma areas and sarcoma components were not separately analyzed by CGH. However, the comparison of cases with predominance of sarcomatous areas with those that mainly showed carcinomatous differentiation revealed a high similarity. In addition, no difference was seen in CGH patterns of carcinosarcomas with different epithelial differentiation (endometrioid vs serous), exemplarily amplification of chromosome 8 was observed in 54% of serous and 62% of endometrioid carcinosarcomas, equally to amplifications on chromosome 20 (72 and 75%). These results show that the two tumor components are genetically closely related and support the “conversion” hypothesis. Furthermore, the CGH patterns observed in our study are almost identical to those of high-grade serous ovarian carcinomas and very similar to CGH patterns of endometrial carcinomas [32].These findings suggest that carcinosarcomas are particularly aggressive variants of conventional ovarian carcinomas.

To gain more detailed information about the two tumor components, we investigated the carcinomatous and sarcomatous tumor component separately using FISH technique. Because amplifications on chromosome 8q and 20q were particularly frequent (about 70%), comparative FISH analysis of the carcinomatous and sarcomatous tumor components focused on these regions. Amplification of znf217 was seen equally in both tumor components, whereas high-level amplifications of c-myc were more prominent in carcinomatous than in sarcomatous tumor areas. Overall, results of CGH and FISH analysis correlated well. C-myc amplifications were less evident in CGH than in FISH analysis. In six cases, amplification of c-myc was seen only by FISH. These results underline the lower sensitivity of CGH for detection of small chromosomal aberrations.

In agreement with molecular–genetic analyses of other investigators [1, 21, 25, 30], we were able to demonstrate that the majority of the tumors harbored a substantial number of gene amplifications in the carcinomatous tumor area indicating a high level of chromosomal instability in these areas.

Separate assessment of the proliferation activity of both tumor components by Ki67 immunohistochemistry are in agreement with this hypothesis. A higher proliferation index was revealed in the carcinomatous areas, whereas the sarcomatous component showed inhomogeneous and focal expression of Ki67 in tumor cells. Other investigators have also demonstrated a higher proliferation index and apoptotic activity in the carcinomatous than in the sarcomatous component [31, 39]. In agreement with this view, it is noteworthy that metastases of carcinosarcomas often only contain carcinomatous tissue. Furthermore, it has been shown by immunohistochemistry of the vascular endothelial growth factor that angiogenic activity is also more pronounced in the carcinomatous areas [11]. These results also suggest that the carcinomatous element is the biologically dominant one in these mixed tumors.

In summary, molecular cytogenetic analysis of 30 cases of ovarian and uterine carcinosarcomas revealed a high level of homology of chromosomal aberrations between the different tumor components suggesting a monoclonal origin of these tumors. However, the carcinomatous element seems to be the driving force of tumorigenesis. The sarcomatous tumor component is either derived from the carcinoma by transdifferentiation or from a common stem cell undergoing divergent differentiation. For confirmation of these results, additional molecular–genetic analyses will be needed.

Furthermore, the striking similarity of molecular genetic aberrations of carcinosarcomas and high-grade ovarian carcinomas supports the notion that similar treatment protocols for adjuvant therapy should be used for these groups of patients.

Acknowledgments

The authors thank Beate Luthardt for excellent technical assistance.

Copyright information

© Springer-Verlag 2007

Authors and Affiliations

  • Alexander Schipf
    • 1
  • Doris Mayr
    • 2
  • Thomas Kirchner
    • 2
  • Joachim Diebold
    • 1
  1. 1.Institute of PathologyKSL LuzernLuzern 16Switzerland
  2. 2.Institute of PathologyLMUMunichGermany

Personalised recommendations