Frequent gene dosage alterations in stromal cells of epithelial ovarian carcinomas

Abstract

Stromal cells are an active and integral part of epithelial neoplasms. We have previously observed allelic imbalance on chromosome 3p21 in both stromal and epithelial cells of ovarian tumors. This study was designed to explore gene dosage alterations throughout human chromosomes from stromal and epithelial cells of epithelial ovarian carcinomas. Thirteen stromal and 24 epithelial samples, microdissected from epithelial ovarian carcinomas, were analyzed using multiplex ligation-dependent probe amplification technique. Analysis covered 110 cancer related genes. Frequent genetic alterations were detected both in the stroma and epithelium of ovarian carcinomas. The mean number of altered genes per tumor was 10.8 in stroma and 23.6 in epithelium. In the stroma, the mean number of gains was 6.6 and of losses 4.2 and in the epithelium 13.7 and 9.9. The high number of changes associated with advanced tumor stage (p = 0.035) and death due to ovarian cancer (p = 0.032). The most frequent alteration was the deletion of the deleted in colorectal carcinoma (DCC) on chromosome 18q21.3 in 62% of samples. Loss of DCC was related to endometrioid subtype (p = 0.033). Large chromosomal aberrations were detected on the basis of alterations in adjacent genes. Most importantly, 38 genes showed similar genetic alterations (gain–gain or loss–loss) in stromal and epithelial compartments of 11 tumor pairs. Thus, frequent genetic alterations in stromal cells of epithelial ovarian carcinomas resembled those of malignant epithelial cells and may indicate a common precursor cell type. Epithelial–mesenchymal transition may generate transformed cancer cells and modify the tumor microenvironment with distinct properties. © 2006 Wiley-Liss, Inc.

It is well established that interrelationship between the tumor epithelium and its microenvironment play a crucial role in tumorigenesis.1, 2, 3 Gene expression profiles of each cell type, composing normal breast tissue and invasive breast carcinomas, have been confirmed to be dramatically different.4 Maffini et al.5 have shown that alterations in stromal cells exposed to carcinogen are necessary and sufficient for the development of epithelial mammary carcinoma. Genetic defects in stromal cells can also stimulate the development of epithelial tumors.6 We have previously found that in epithelial ovarian cancer on chromosome 3p21 allelic imbalance is found in 52–80% of the stromal cells and in 60–87% of the epithelial tumor cells.7 Genetic alterations in stromal cells adjacent to malignant tumors have also been detected in breast and colon cancers.8, 9, 10, 11, 12

The purpose of this study was to define the extent of genetic alterations in stromal cells of epithelial ovarian carcinomas, by comparing the alterations between stromal and epithelial cells and by this add to the knowledge of the relationship between stromal and epithelial cells. We analyzed the gene dosage of 110 different cancer associated genes, covering all human chromosomes from stromal and epithelial compartments of 26 ovarian carcinoma samples, using multiplex ligation-dependent probe amplification assay (MLPA).13 Detected gains and losses in the stroma and epithelium were compared with the clinicopathological features of the tumors.

Material and methods

Samples

The material consisted of 13 stromal and 24 epithelial samples from 26 epithelial ovarian carcinomas. Both the stromal and epithelial areas could be analyzed from 11 carcinomas. These 11 pairs were used to research the possible occurrence of epithelial–mesenchymal transition (EMT). Carcinomas were collected from a series of 316 patients, diagnosed and treated for epithelial ovarian cancer at Kuopio University Hospital and Jyväskylä Central Hospital, Finland, between 1976 and 1992.14 The samples for the analyses were selected on the basis of available qualified archived material. The patients were followed up until September 1996. All tumors were staged according to the International Federation of Gynecologists and Obstetrics (FIGO) standards, and histological typing and grading was performed according to the World Health Organization (WHO) classification (Table I).15 Normal uterine tissue from the same patients was used as a control.

Table I. Clinicopathological Characteristics of the Patients

Variable	n = 26
1 Values in square brackets indicate ranges. 2 Values in parentheses indicate percentages. 3 Miscellaneous tumors contained 3 mucinous, 2 clear cell, 3 mixed epithelial and 1 unclassified epithelial tumors. 4 FIGO, International Federation of Gynecologists and Obstetrics. 5 In addition, there were 10 cases in which patient's chemotherapy response was not verified. 6 Death, due to other reason, 3 cases.
Years of diagnosis, years	1977–1992
Median age at diagnosis, years	63 [34–82]1
Histologic subtype
Serous	7 (27)2
Endometrioid	10 (38)
Miscellaneous3	9 (35)
Histologic grade
1	3 (12)
2	10 (38)
3	13 (50)
FIGO4 stage
I	10 (38)
II	2 (8)
III	12 (46)
IV	2 (8)
Chemotherapy response5
Yes	9 (56)
No	7 (44)
Patient status6
Dead, ovarian cancer	14 (61)
Alive	9 (39)
Median follow-up time, months	35 [4–280]
Median overall survival, months	33 [4–184]

Tissue microdissection

Five sequential 10-μm thick sections from the formalin-fixed and paraffin-embedded tumor material were microdissected manually by an experienced pathologist, using a 24-G needle, under a microscope. Epithelial tumor cells and peritumoral stromal cells were microdissected separately (Fig. 1). Epithelial areas were confirmed to contain at least 70% tumor cells. DNA was extracted from the specimens with the proteinase-K–phenol–chloroform method, following standard protocols.16

Figure 1. Manual microdissection from H&E-stained section from a serous ovarian cancer specimen. Epithelial (+) and stromal (*) tumor areas (a) before and (b) after dissection. Ten-fold magnification.

Multiplex ligation-dependent probe amplification

Gene dosage of 110 different genes was analyzed using MLPA.13 Human chromosomal aberration test kits 1, 2 and 3 were used according to manufacturer's instructions (MRC Holland, Amsterdam, The Netherlands). The 3 human chromosomal aberration test kits covered altogether 112 genes, but 2 were rejected due to low intensity (TIMP2, NCOA3). In brief, after denaturation, DNA probe mix was added and heated at 95°C for a minute and incubated at 60°C for 16 h. Ligation was performed at 54°C for 15 min, and heat-stable ligase was inactivated at 98°C. PCR was carried out for 40 cycles (20 s at 95°C, 30 s at 60°C, 60 s at 72°C and at the end at 72°C for 20 min) with FAM-labeled primers. Amplification products were analyzed utilizing ABI Prism 310 genetic analyzer (Applied Biosystems, Foster City, CA).

In the data normalization procedure, the relative peak height for each probe was obtained as fractions of the total sum of peak heights in a certain sample. The fraction of each probe was divided by the average peak fractions of the corresponding probe in control samples. Highly amplified genes (normalized copy number >2) were left out when calculating the total sum of peak heights, and the peak heights were renormalized. A difference was interpreted, significant only when the ratio was less than 0.66 (loss) or higher than 1.34 (gain). The normalized copy numbers of all normal control samples were between 0.66 and 1.34.

Statistical analyses

Significance levels for an association between MLPA results and the clinicopathological features were computed by Fisher's exact tests, using SPSS 11.5 software. p values <0.05 were considered significant.

Results

The stromal samples from epithelial ovarian carcinomas possessed extensive genetic aberrations. In all analyzed 13 stromal and 24 epithelial tissues, quantitative copy number changes were detected. The mean number of quantitatively altered genes per tumor was 10.8 in stroma compared with 23.6 in epithelium, when 110 different genes were analyzed (Table II). The number of altered genes per tumor ranged from 1 to 43. We did not find any changes in normal uterine tissue. The most frequent alteration was loss of DCC on chromosome 18q21 (Table III). Altogether, copy number changes were detected in 67 (61%) out of 110 genes in stromal samples and in 106 (96%) genes in epithelial samples.

Table II. The Number of Genetic Alterations1 per Tumor in Stromal and Epithelial Compartments

	Stroma (n = 13)		Epithelium (n = 24)
	Mean n	Range	Mean n	Range
1 Altogether, 110 different genes were analyzed.
Gains per tumor	6.6	1–15	13.7	1–23
Losses per tumor	4.2	0–12	9.9	0–20
Changes per tumor	10.8	2–23	23.6	1–43

Table III. The Most Frequently Altered Genes in Stromal (n = 13) and Epithelial (n = 24) Samples

Gene(s)	Changes, n
1 Values in parentheses indicate percentages.
Gain
Stroma
CDKN2B, SCYA3	5/13 (38)1
GF3, TP53	4/13 (31)
TNFRSF1B, CTPS, RECQL4, HRAS, TNFRSF1A, ERBB2	3/13 (23)
Epithelium
PTP4A3, PTPN1	13/24 (54)
MYC	12/24 (50)
IL12A	11/24 (46)
Loss
Stroma
DCC	8/13 (62)
CTPS, ING1	3/13 (23)
CASP2, PTK2, IGF1R, CRK, BCL2	2/13 (15)
Epithelium
DCC	15/24 (63)
IL2	10/24 (42)
PPEF1, PDCD8	9/24 (38)

Altered genes could be found either in both stroma and epithelium or in either one of the compartments within tumor pairs (Table IV). Moreover, 38 genes were affected similarly (gain–gain or loss–loss) in both compartments of epithelial–stromal pairs (Table V). In the case of DCC, 10 of 11 tumor pairs showed similar results in stromal and epithelial compartments. Dissimilar changes were detected in 4 genes.

Table IV. Genetic Alterations in 11 Stromal–Epithelial Tumor Pairs

Tumor	Number of genes changed only in stroma, n	Number of genes changed only in epithelium, n	Number of genes changed similarly in stroma and epithelium, n	Number of genes changed dissimilarly1 in stroma vs. epithelium, n
1 Gain vs. loss or loss vs. gain.
1	1	13	1	0
2	13	17	6	1
3	9	11	11	1
4	5	14	4	0
5	1	33	4	0
6	7	9	16	0
7	3	3	3	0
8	2	21	6	0
9	2	17	1	0
10	1	24	0	1
11	2	36	3	1

Table V. The Genes Changed Similarly (Gain–Gain or Loss–Loss) in Stromal–Epithelial Pairs of 11 Tumors

Genes	Number of gains in stromal–epithelial pair	Number of losses in stromal–epithelial pair	Number of dissimilar changes in stromal–epithelial pair1
1 Gain vs. loss or loss vs. gain.
DCC	0	7	0
SCYA3	3	0	0
c-MYC	2	0	0
CCND2	2	0	0
TNFRSF1A	2	0	0
TP53	2	0	1
BCL2L1	2	0	0
PTK2	0	2	0
ING1	0	2	0
CTPS	1	1	1
ERBB2	1	1	0
CFLAR	1	0	0
IL12A	1	0	0
Hs.222808	1	0	0
HLA-G	1	0	0
LTA	1	0	0
BAK1	1	0	0
VEGF	1	0	0
RECQL4	1	0	0
CDKN2B	1	0	0
FGF3	1	0	0
EMS1	1	0	0
TNFRSF7	1	0	0
LRMP	1	0	0
MVP	1	0	0
CDH2	1	0	0
CDKN2D	1	0	1
THBD	1	0	0
MIF	1	0	0
CASP6	0	1	0
FGFR1	0	1	0
CRK	0	1	0
PMAIP1	0	1	0
BCL2	0	1	0
PPEF1	0	1	0
AR	0	1	0
PDCD8	0	1	0
FRM2	0	1	0
CASP2	0	0	1

Several affected chromosomal regions were found in ovarian tumors by MLPA method (Fig. 2). Loss of DCC on chromosome 18q21 occurred in 62% of the tumors both in epithelium and stroma. IL2 on chromosome 4q was deleted in 42% of tumor epithelium. Very frequent gain of MYC (50%) and adjacent genes was detected on chromosome 8q in epithelial compartments. There was another common gain region, telomeric to MYC on 8q between PTP4A3 and RECQL4 (54–42%). The genes between MYC region and PTP4A3-RECQL4 region had a normal copy number. On chromosome 20q, BCL2L1 (42%) and PTPN1 (54%) had frequent gains in epithelium. On chromosomal region 3q25-q29, frequent gains in epithelial compartments of the tumors were seen. The most frequently altered gene in this region was IL12A (46%). On chromosomal region 11q13 gain of FGF3 gene (42%) was the most common alteration in epithelium. On chromosome 17, TP53 gene locus showed an increased copy number in 38% of epithelial and 31% of stromal samples and decreased in 21% epithelial and 4% stromal samples. ERBB2 gain was detected in 33% of epithelial and 4% of stromal samples. The same type of change was often seen in adjacent genes, indicating larger chromosomal aberrations. This could be seen both in stromal and epithelial compartments of the tumors, such as the loss of X chromosome detected in parallel with all X chromosome targeted probes.

Figure 2. MLPA results of all chromosomes from 26 analyzed carcinomas. Gains are marked on the right side of chromosomes and losses on the left side. Epithelium: black boxes, stroma: white boxes. The length of the bar corresponds with the frequency of copy number change. Large chromosomal aberrations are marked with vertical lines, and the most commonly altered genes in these regions in epithelium are underlined.

Statistical analyses were performed to determine whether some particular changes in the genes were associated with common factors in clinicopathological determinants (Table I). The tumors with higher mean number of changes (23.6) in epithelium were associated with advanced tumor stage III/IV (p = 0.035) and death because of ovarian cancer (p = 0.032). Loss of DCC was associated with endometrioid histological subtype (p = 0.033). Copy number changes of BAK1, LTA, IL12A and RB1 genes were related to advanced tumor stage (III/IV), with p-values 0.013, 0.033, 0.036 and 0.047, respectively. IL12A change (p = 0.008) and gain of TNFRSF1A (p = 0.018) were associated with ovarian cancer death. An alteration in TP53 was related to histologic grade 3 (p = 0.015) and normal stromal ERBB2 was related to histologic grade 2/3 (p = 0.038). No further relations to clinicopathological features could be found.

Discussion

In this study, we show that the genetic alterations found in the epithelial malignant cells of ovarian carcinomas are also present in the cells of surrounding mesenchymal stroma, in a substantial portion of the studied tumors. These irreversible genetic alterations resembled those of malignant epithelial cells, and thus indicate that a proportion of stromal cells are transformed toward malignancy and thus should be considered as an active part of the tumor. Frequent genetic alterations are also quite likely to affect the function of stromal cells and to disturb the normal epithelial–stromal interactions and tissue homeostasis. Stromal cell contribution to tumorigenesis may provide new diagnostic markers and therapeutic targets.

Recently published data suggest that EMT generates a tumor stroma containing mesenchymally transformed epithelial cancer cells.17, 18 EMT is a well-described process in morphogenesis, injured tissues and tumor progression, and it associates with dedifferentiation and invasion steps.19, 20, 21 According to our results, similar genetic alterations in epithelial and stromal cells of ovarian carcinomas might not have developed independently in different tissues, but instead alterations may have occurred in 1 tissue type followed by EMT (Table V). Similar genetic alterations (allelic imbalance or mutations) in both stromal and epithelial compartments of the tumors have been observed in the present and previous studies.7, 8, 9, 10, 11, 12 Additionally, a nonrandom X-chromosome inactivation pattern shared with epithelial tumor cells has been found in a mesenchymal-like cell line derived from breast carcinoma.22 However, Fukino et al.12 have found that loss of heterozygosity patterns between epithelium and stroma differentiate significantly, and therefore suggested that epithelial and stromal cells would originate along 2 independent paths. A possible effect other than EMT leading to similar genetic alterations could be selective pressure targeted to epithelial and stromal cells in the tumor microenvironment.12 According to Allinen et al.,4 there is only a low probability that 2 adjacent tissues would acquire parallel genetic alterations. Our data show some discordance in the number and type of genetic alterations between stroma and epithelium. We believe this is the result of intratumoral genetic heterogeneity, tumor progression and acquisition of new alterations after EMT. In line with our results, a recent study of BRCA1 mutation-positive breast cancer cases showed shared cluster of allelic imbalance in epithelium and stroma, and evidenced for EMT.23 Although our results support EMT, forthcoming studies, including in vivo studies, will show how important role EMT has in tumorigenesis.

DCC gene belongs to a family of dependence receptors and locates on chromosomal region 18q21, which is one of the most commonly deleted regions in epithelial ovarian cancer detected by loss of heterozygosity24 and comparative genomic hybridization analyses.25 In the present study, the existence of a common precursor cell of stroma and epithelium is supported by the notion that loss of the DCC locus, the most frequently observed alteration in this study, was detected either in both epithelial and stromal compartments of the tumor or in none of them. Loss of the DCC locus could be an early step in epithelial components of endometrioid ovarian tumorigenesis, since the loss was not associated with tumor histopathological grade or FIGO stage, or our study was too small to demonstrate an association.

Widespread DNA copy number alterations observed in this study indicate alterations in several genes that are not critically involved in carcinogenesis, but locate close to cancer-related tumor suppressor genes or oncogenes targeted by a loss or gain. Therefore, large chromosomal aberrations can have a direct effect on expression levels of these adjacent bystander genes26, 27 and, in expression studies, the possibility of large chromosomal aberrations as a cause of altered expression should be examined. On the other hand, alterations of multiple adjacent genes may provide some growth advantage that individually altered genes do not have.

This study shows that MLPA is an efficient and reliable method for multiple parallel analyses of quantitative genetic alterations from small amounts of fragmented tumor DNA.28 In the epithelial compartment, the well-known ovarian cancer oncogenes and tumor suppressor genes, such as MYC, ERBB2 and TP53, showed similar magnitude of changes, as reported in the literature.29, 30, 31 Control male samples showed always the presence of single X and Y chromosomes. In addition, there were no changes in normal uterine tissues. There is a possibility that observed alterations in stromal cells could be caused by contaminating epithelial cells in stromal tissue. This is quite unlikely due to several reasons. First, in our previous study,7 with the same carcinoma material, we took advantage of both laser capture microdissection and manual microdissection, and the results of allelic imbalance analyses were equal. Second, in the present work, the normalized values, from epithelial and stromal compartments of the same tumor, were mostly multiples of the exact copy number of the gene, such as 2, 3 or 4, which would not be the case in the presence of contamination.

In conclusion, we show that stromal cells of epithelial ovarian carcinomas possess multiple copy number changes, highly similar to alterations detected in epithelial cells. Tumor stroma seems to contain cells of epithelial origin that are generated by EMT, and further, this transformed tumor stroma may have an effect on epithelial–stromal cell interactions and tumorigenesis. Loss of DCC on chromosome 18q21 was the most common alteration, which occurred equivalently in stromal and epithelial compartments of tumor. In addition, MLPA is shown to be a fast and reliable method to screen quantitative changes in multiple genes, using limited amount of degraded DNA obtained from paraffin-embedded material.

Acknowledgements

We thank Senior Lecturer Raija Tammi and Professor Markku Tammi for critical review of the manuscript.