Leiomyosarcomas (LMS) are a common subtype of soft tissue sarcoma. The molecular causes of the disease remain unclear.
Leiomyosarcomas (LMS) are a common subtype of soft tissue sarcoma. The molecular causes of the disease remain unclear.
In the current study, gene expression in LMS, leiomyomas, and normal myometrium was examined. RNA was prepared and gene expression was determined using microarray analysis arrays containing approximately 12,000 known genes and 48,000 expressed squence tags (ESTs).
A number of genes were found to be differentially expressed in these sample sets, and six genes including cyclin-dependent kinase inhibitor 2A, diaphanous (Drosophila homolog) 3, doublecortin, calpain 6, interleukin-17B, and proteolipid 1 were found to be overexpressed in LMS compared with normal myometrium and 18 other tissues. Sets of genes were identified whose expression could be used to cluster samples with either LMS, leiomyomas, or normal myometrium.
The authors concluded that differences in gene expression can be detected between LMS and leiomyomas, normal myometrium, and other tissues, and that these changes in gene expression may yield clues with regard to the pathophysiology of leiomyosarcoma. Cancer 2003;98:1029–38. © 2003 American Cancer Society.
Leiomyosarcoma (LMS) forms a common subtype of soft tissue sarcoma that is difficult to treat when metastatic.1–3 Malignant transformation generally is believed to be associated with changes in the expression of a number of genes, and this alteration in gene expression is considered to be critical to the development of the malignant phenotype. In many cases, the progression to malignant transformation is associated with the sequential acquisition of multiple mutations. A number of genetic changes have been observed in LMS, and commonly include alterations in p53 and MDM2 expression4 and a loss of gamma-smooth muscle isoactin gene expression.5 LMS has been reported to typically have a highly aneuploid karyotype.6
Recent advances in knowledge of the human genome and the ability to perform expression profiling have provided new capabilities to analyze tumors at a molecular level.7 Recently, we compared gene expression in leiomyomas and myometrium by gene array analysis and identified four gene fragments (doublecortin, calpain 6, interleukin-17B, and proteolipid 1) that were differentially expressed in leiomyoma compared with myometrium and other tissues. To extend our earlier study and to better understand the biology of leiomyosarcoma, expression of approximately 60,000 genes/expressed sequence tags (ESTs) in LMS, uterine leiomyomas, and normal myometrium was determined by the Affymetrix microarray technique (Affymetrix, Inc, Santa Clara, CA) and differences in gene expression were analyzed. The expression of genes of interest was then examined in 250 other samples obtained from 18 different types of tissue. We concluded that differences in gene expression may help to characterize LMS, yield clues regarding the pathophysiology of this common category of tumor, and identify potential targets for therapy.
Tissue from 4 patients with uterine LMS (age range, 51–61 years), 4 patients with nonuterine LMS (age range, 52–73 years), 19 patients with uterine leiomyoma (age range, 26–87 years; median age of 55 years), 46 patients with normal myometrium (age range, 27–84 years; median age of 52 years), and 250 other samples from 18 different types of tissues (including 12 normal adipose tissue samples, 7 samples of normal cervix, 24 samples of normal colon, 11 samples of normal kidney, 12 samples of normal liver, 24 samples of normal lung, 12 samples of normal skeletal muscle, 15 samples of normal ovary, 9 samples of normal skin, 8 samples of normal small intestine, 55 samples of normal thymus, 11 samples of tonsils with lymphoid hyperplasia, 7 colon adenocarcinoma samples, 7 lung adenocarcinoma samples, 11 papillary ovarian serous adenocarcinoma samples, 8 renal cell carcinoma samples, 9 squamous carcinoma of the lung samples, and 8 samples of gallbladder with chronic inflammation) were obtained from the Tissue Procurement Facility of the University of Minnesota. Samples were obtained using protocols approved by the University of Minnesota Institutional Review Board. Tumor and normal samples were identified, dissected, and snap frozen in liquid nitrogen within 30 minutes of removal from the patient. Tissue sections of each sample were prepared before freezing and were examined by light microscopy after hematoxylin and eosin staining to confirm the pathologic nature of the sample. None of the samples was necrotic.
RNA was prepared and gene expression was determined at Gene Logic Inc. (Gaithersburg, MD) using Affymetrix GeneChip® U_95 arrays containing approximately 12,000 known genes and 48,000 ESTs. Gene expression analysis was performed with the Gene Logic Gene Express® Software System using the Gene Logic normalization algorithm. A fold change analysis was performed in which the ratio of the geometric means of the expression intensities of the relevant gene fragments were computed and the ratio was reported as the fold change (up or down). Ninety-five percent confidence intervals (95% CIs) and P values on the fold change also were calculated using a two-sided Welch modified two-sample Student t test. Differences were considered significant when P ≤ 0.05. e-Northern™ analyses were performed using the Gene Logic Gene Express® Software System. Clustering was performed using the Eisen clustering software (Eisen Lab, University of California at Berkeley, Berkeley, CA) and viewed using the TreeView software (available from URL: http://rana.lbl.gov).
Gene expression using the Affymetrix GeneChip® U_95 chip set was performed on all samples. Approximately 5000 of the 60,000 gene fragments were present in all 4 samples in the set of uterine LMS. This number did not appear to vary greatly if three or four samples from the set were included in the analysis (Fig. 1, top panels). As with the set of uterine LMS, approximately 5000 gene fragments were present in all the samples of the nonuterine LMS set, and little variation was noted when 3 or 4 samples were included in the analysis (Fig. 1, middle panels). We recently reported that six samples appeared to be sufficient to achieve little change in the number of gene fragments expressed in all members of the sample set as further samples are added for the sets of leiomyomas and normal myometrium samples.8 When the uterine and nonuterine LMS were pooled, approximately 3500 gene fragments were present in all the samples of the combined set, and little variation was observed when ≥ 5 samples were included in the analysis (Fig. 1, bottom panels). This suggests that LMS does vary somewhat depending on the location of origin, yet the majority of the gene fragments expressed are similar. We therefore combined the two sets of LMS into one set of eight samples for further analysis.
The relative intensities of gene expression in the sample sets were compared. A total of 585 gene fragments were expressed at ≥ 4 fold different levels in the LMS set compared with the set of normal myometrium (Table 1). The known genes most overexpressed in LMS compared with normal myometrium are shown in Table 2. Osteopontin (also known as secreted phosphoprotein 1 and bone sialoprotein 1), pituitary tumor-transforming 1 (PTT-1), and ubiquitin-conjugating enzyme E2C (E2C) were expressed > 10-fold more in the set of LMS compared with normal myometrium. CDC2-associated protein CKS2 (CKS2), cellular retinoic acid-binding protein 2, cyclin-dependent kinase inhibitor 2A (CDKIN2A), diaphanous (Drosophila homolog) 3, forkhead box M1B, interleukin (IL)-17B, popeye protein 2, protein regulator of cytokinesis 1 (PRC-1), suppression of tumorigenicity 5 (ST5), and topoisomerase II-α were expressed 5-fold to 10-fold more in LMS compared with normal myometrium.
|Fold change range||Down in LMS||Up in LMS||Total up or down in LMS|
|CDC2-associated protein CKS2|
|Cellular retinoic acid-binding protein 2|
|Cyclin-dependent kinase inhibitor 2A (CDKIN2A, p16)|
|Diaphanous (Drosophila homolog) 3|
|Forkhead box M1B|
|Osteopontin (bone sialoprotein 1, secreted phosphoprotein 1)|
|Pituitary tumor-transforming 1 (PTT-1)|
|Popeye protein 2|
|Protein regulator of cytokinesis 1 (PRC-1)|
|Suppression of tumorigenicity 5 (ST5)|
|Ubiquitin-conjugating enzyme E2C|
A total of 425 gene fragments were expressed at ≥ 4-fold different levels in the LMS set compared with the leiomyoma set (Table 3). Of the known genes, osteopontin, E2C, PTT-1, CDKIN2A, and forkhead box M1 were expressed at 5-fold to 20-fold more in the set of LMS compared with leiomyomas. Thus, many of those genes that were overexpressed in LMS compared with normal myometrium also were overexpressed in LMS compared with leiomyomas.
|Fold change range||Down in LMS||Up in LMS||Total up or down in LMS|
We previously reported that calpain 6, doublecortin, IL-17B, and proteolipid-1 were overexpressed in uterine leiomyomas compared with normal myometrium, and were either not expressed or expressed at very low levels in 18 other tissue types.8 Therefore, the expression of these genes also was examined in the LMS samples. Doublecortin was expressed in two of four uterine LMS samples and in one of four nonuterine LMS samples (Fig. 2). Calpain 6 was expressed in two of four uterine LMS samples and two of four nonuterine LMS samples (data not shown). IL-17B also was found to be overexpressed in the LMS set compared with normal myometrium in the fold change analysis as described earlier, and was expressed in three of four uterine LMS samples and two of four nonuterine LMS samples (data not shown). Proteolipid-1 was expressed in two of four uterine LMS samples and in one of four nonuterine LMS samples (data not shown).
Some gene fragments were underexpressed in LMS compared with normal myometrium (Table 1). Among the known genes that were expressed more than fivefold less in LMS compared with myometrium were alcohol dehydrogenase 1A-α polypeptide, alcohol dehydrogenase 1B-β polypeptide, insulin-like growth factor 1, c-jun, c-fos, and TU3A. A representative experiment illustrating the low levels of gene expression for LMS compared with normal myometrium is shown for alcohol dehydrogenase 1A-α polypeptide (Fig. 3).
All known genes that were overexpressed twofold to fourfold more in LMS compared with myometrium were examined by Contrast Analysis™ to identify those genes whose expression was specific to LMS compared with muscle, adipose, kidney, and liver, and these genes then were combined with the set of all known genes that were overexpressed more than fourfold in LMS compared with myometrium to generate another set of gene fragments. To explore the specificity of expression of these gene fragments further, the expression of these gene fragments in 18 other normal and diseased tissues was examined by the Gene Logic e-Northern™ analysis software from the Gene Express® Software System. This analysis provides a graphic representation of the level of gene expression in each sample of a sample set. Two genes, CDKIN2A and diaphanous (Drosophila homolog) 3, were found to be overexpressed specifically and significantly in LMS compared with normal myometrium and uterine leiomyomas, and were expressed only at very low levels in the other 18 types of tissues examined. CDKIN2A was selectively overexpressed in LMS among the normal tissues examined (Fig. 4). Expression in some ovarian, lung, and colon tumors also was observed. Diaphanous (Drosophila homolog) 3 was found to be overexpressed in five of seven LMS samples and was expressed at very low levels or not at all in the other tissues examined (data not shown).
Several other genes were overexpressed in many of the LMS samples compared with the other tissues examined, but also were found to be significantly expressed in some other tissues. For example, PRC1 was overexpressed in four of eight LMS samples and also was expressed in thymus and ovarian carcinoma samples compared with the other tissues examined. PTT-1 was overexpressed in LMS compared with the majority of normal tissues, but also was expressed in normal colon, thymus, and tonsil tissue, as well as in carcinomas of the colon, lung, ovary, and kidney.
Cellular retinoic acid-binding protein 2 was found to be overexpressed in leiomyomas and LMS compared with the majority of normal tissues, and several members of the ubiquination pathway were found to be overexpressed in LMS samples compared with most normal tissues (except that of the tonsil and thymus and several tumors), although they also were expressed at some level in the majority of tissues. Many of the other up-regulated genes also were expressed at similar levels in a variety of other tissues, and therefore were not specific to LMS.
The sets of gene fragments expressed at twofold more or less in the LMS sample set compared with the normal myometrium sample set, the set of gene fragments that were expressed twofold more or less in the LMS sample set compared with the leiomyoma sample set, and the set of gene fragments expressed twofold more or less in the leiomyoma sample set compared with the myometrium sample set were combined and subjected to Contrast Analysis using the Gene Logic Gene Express® Software System to identify those genes whose expression was most significantly different between the LMS and the other two sample sets. Using this analysis, a set of 540 gene fragments was chosen for clustering. Clustering was performed using the Eisen clustering software, Cluster. When clustering was performed using this restricted set of gene fragments, all the LMS and leiomyoma samples clustered separately from the normal myometrium samples (Fig. 5). The significance of the subclusters within the leiomyoma cluster is unclear. Leiomyomas appear microscopically very similar to normal myometrium, but there is heterogeneity among leiomyomas. It should be noted that the left-right position around a node on the tree has no significance; rather, it is the length of the line to the node that indicates the similarity of the gene expression pattern to the next cluster.
Although the uterine and nonuterine LMS sets each had only four samples, these sets were compared by fold change and Contrast Analysis™ as described earlier, and a number of differences in gene expression were observed. However, no genes were found to be uniquely expressed in either uterine LMS or nonuterine LMS samples.
In the current study, the expression of approximately 60,000 gene fragments in LMS was examined. A large number of genes were found to be expressed differentially in LMS compared with normal myometrium and leiomyoma samples. The genes expressed at a level of twofold or more in LMS were analyzed for their expression in a variety of other normal and diseased tissues. Two genes, CDKIN2A and diaphanous (Drosophila homolog) 3, were found to be overexpressed in LMS only compared with the 235 normal tissues examined. Other genes were found to be overexpressed in LMS compared with leiomyoma and normal myometrium, but also were expressed in a variety of other tissues.
CDKIN2A, also known as p16, was selectively overexpressed in LMS compared with nearly all the other tissue sets. CDKIN2A binds cyclin-dependent kinases CDK4 and CDK6, thus inhibiting the ability of CDK4 to interact with cyclin D and phosphorylation of the RB protein by CDK4, leading to G1 cell cycle arrest.9–11 Thus, CDKIN2A can act as a tumor suppressor, with mutation or loss of CDKIN2A or RB and overexpression of D cyclins having similar effects on cell cycle progression.11, 12 Deletion or mutation of CDKIN2A in tumor and tumor cell lines has been reported.11, 13 It is not known whether CDKIN2A in the LMS samples cited herein contain mutations. Some CDKIN2A mutations predispose to melanoma.11, 14, 15 However, a study of colorectal carcinoma found overexpression of CDKIN2A in 14 of 17 tumors compared with adjacent normal tissue, and no mutations in the gene were found, suggesting that overexpression rather than inactivation of the gene may play a role in the pathogenesis of colorectal carcinoma.16 To our knowledge, little is known regarding the function of diaphanous (Drosophila homo-log) 3.
Four genes previously reported to be specifically overexpressed in leiomyoma compared with myometrium8 (doublecortin [doublecortex], calpain 6, IL-7B, and proteolipid protein 1) also were found in the current study to be overexpressed in LMS compared with myometrium, and also were either not expressed or expressed at very low levels in the other 250 tissue samples examined. Doublecortin originally was identified in studies of lissencephaly, in which mutations in the doublecortin gene lead to cortical neuronal migration syndromes.17 In human fetal tissues, doublecortin mRNA is expressed exclusively in the nervous system, where it tends to localize in areas of ongoing neuronal migration.18–20 Doublecortin binds tubulin17 and also interacts with the adaptor complexes AP-1 and AP-2, which are involved in clathrin-dependent protein sorting, and therefore may play a role in protein sorting or vessicular trafficking.21 Many functions of calpains have been proposed, including regulation of the cell cycle, apoptosis, and cell adhesion and motility.22, 23 It is interesting to note that the recently described doublecortin-like kinase is comprised of a doublecortin-like domain that localizes to microtubules and a C-terminal serine-threonine kinase domain. Doublecortin-like kinase is a substrate for calpain, which cleaves the kinase domain from the microtubule-binding site.22 Doublecortin-like kinase was not found to be overexpressed in LMS in the current study. It is possible that the ratio of doublecortin to doublecortin-like kinase could be functionally significant. IL-17 is a cytokine that may be important in the initiation or maintenance of the proinflammatory response. IL-17B is related to IL-17 but has been detected in a variety of adult tissues, although to our knowledge not in T-cells.24, 25 Finally, mutations and/or gene duplication of proteolipid protein 1, or lipophilin, identified in studies of Pelizaeus-Merzbacher disease, are reported to lead to multiple neurologic syndromes.11
Other genes found to be overexpressed in LMS compared with the majority of other tissues in the current study included osteopontin, PTT-1, PRC-1, CKS2, cellular retinoic acid-binding protein 2, forkhead box M1B, popeye protein 2, ST5, and topoisomerase II-α. Osteopontin, which is produced by osteoclasts stimulated by calcitriol, binds hydroxyapatite and anchors osteoclasts to the bone matrix, where it is the major phosphorylated glycoprotein of bone.11 Osteopontin also forms the proteinaceous matrix of urinary stones,11, 26 is a component of normal elastic fibers in skin and aorta,11 and is an important cytokine for type I immune responses mediated by macrophages in mice.11 The osteopontin gene is a target gene of p53 and osteopontin recently was found to be up-regulated by DNA damage-induced p53 activity.11, 27 Changes in p53 have been commonly observed in LMS.4
PTT-1 originally was isolated from pituitary tumors28 and reportedly is overexpressed in various tumors and hematopoietic malignancies.28–30 PTT-1 regulates sister chromatid separation, and thus normal PTT-1 function may contribute to the maintenance of euploidy.31, 32 In addition, conditioned medium from transfectants that overexpress PTT-1 induced proliferation and tube formation of human umbilical vein endothelial cells, and PTT-1 can induce an angiogenic phenotype in in vitro and in vivo angiogenesis models, possibly via bFGF.28, 33
Many of the other genes found to be overexpressed in LMS play important roles in cell cycle regulation or cell division, including cyclins and cyclin-dependent kinases. Cyclin-dependent kinases, activated by complex formation with cyclins, are important regulators of the cell cycle. PRC-1 protein, a substrate for several cyclin-dependent kinases, is expressed at high levels during the S-phase and the G2/M-phase, but is expressed at much lower levels in the G1 phase. Microinjection of antibodies to PRC-1 blocks cellular cleavage but not nuclear division, indicating a role in cytokinesis.34 The cell cycle regulating protein CKS2 binds cyclin-dependent kinases and is important in their function, thus participating in control of the cell cycle. Several gene fragments associated with transcription, including some involved in cyclin regulation, also were found to be overexpressed in LMS compared with other tissues. For example, forkhead box M1B is a transcription factor that stimulates expression of both cyclin B1 and cyclin D1 promoters, suggesting that forkhead M1B controls the transcription network of genes important for cell division.11, 35
The data also suggest treatment approaches that may be worthy of study. Several genes encoding members of the ubiquitin pathway were found to be overexpressed in LMS compared with other tissues, including E2C, which was prominently overexpressed in LMS compared with the majority of normal tissues except tonsil and thymus. Many proteins that play important roles in regulating gene expression and cell proliferation are degraded by the ubiquitin-proteasome pathway.36 In this pathway, ubiquitin is covalently linked to the protein, which then targets the protein to the 26S proteasome, an ATP-dependent complex, which degrades it.36 Proteasome inhibitors can increase killing of cancer cells, and some proteasome inhibitors currently are being studied in clinical trials. The overexpression of components of the ubiquitin-proteasome pathway in LMS suggests that the study of proteasome inhibitors in these tumors would be of interest. Second, cellular retinoic acid-binding protein 2 was found to be overexpressed in LMS compared with the majority of normal tissues, and overexpression of this gene in leiomyomas compared with myometrium was reported recently.8 In addition, retinoic acid-regulated nuclear matrix-associated protein also was reported to be overexpressed in three of seven LMS samples compared with all normal tissues examined except that of the thymus. Thus, further studies to examine the effects of retinoic acid analogues in selected LMS cases may be warranted. It is interesting to note that topoisomerase II-α was overexpressed in four of eight LMS samples compared with normal tissues except thymus, and this may contribute to the efficacy of etoposide in some cases of LMS.
Several other genes of unknown function were also found to be overexpressed in LMS compared with other tissues. Popeye protein 2 was found to be overexpressed in LMS in the current study, and overexpression of popeye protein 2 in leiomyomas compared with normal myometrium was reported recently.8 The popeye 2 gene is expressed primarily in skeletal and cardiac muscle and has unknown functions, but also is up-regulated in the uterus of pregnant mice.37 Suppressor of tumorigenicity 5 (ST5) also was found to be overexpressed in LMS compared with myometrium in the current study. ST5 was identified as a gene that, when added to HeLa cells, suppresses tumorigenicity.38
In the current study, several genes were found to be underexpressed in LMS, including alcohol dehydrogenase 1A-α polypeptide, alcohol dehydrogenase 1B-β polypeptide, insulin-like growth factor (IGF)-1, c-jun, c-fos, and TU3A. Alcohol dehydrogenase,8 c-fos, and c-jun8, 39, 40 have been reported to be underexpressed in leiomyomas compared with myometrium. c-fos and c-jun form dimeric complexes that bind AP-1-containing sites and thereby regulate gene transcription.40 TU3A was found to be down-regulated in LMS in the current study, and recently was reported to be underexpressed in leiomyomas compared with myometrium.8 TU3A is a gene of unknown function that is widely expressed in normal tissues but found in a region that is commonly deleted in renal cell carcinoma.41 In the current study, both IGF-1 and IGF-2 were found to be underexpressed in LMS compared with myometrium or leiomyomas. IGF-1 mRNA has been reported to be expressed more highly in leiomyomas than in myometrium, independent of the menstrual cycle,42 and IGF-2 mRNA has been reported to be expressed at low levels or to be absent in some LMS samples, and expressed at high levels in other LMS samples.43, 44 Overexpression of IGF-2 in leiomyoma compared with myometrium was reported recently.8
Using the clustering software, Cluster, and the 540 gene fragments selected to examine gene expression in the sets of LMS, leiomyoma, and normal myometrium, all the LMS, leiomyoma, and myometrium samples were found to cluster separately. The significance of the subclusters within the leiomyoma cluster is unclear. Leiomyomas appear microscopically very similar to normal myometrium. In addition, leiomyoma represents a biologically heterogeneous group of tumors, and this heterogeneity may be reflected in the clustering of the leiomyoma samples studied here. Two recent studies have used microarray techniques to examine gene expression in soft tissue sarcomas.45, 46 These studies confirmed the ability to effectively categorize many subtypes of soft tissue sarcoma on the basis of mRNA expression.
An important consideration for studies of gene expression in tumors is the potential for variable gene expression in different parts of the tumor examined. A recent study by Shmulevich et al. regarding gene expression in LMS found little variability among samples obtained from different parts of the tumor compared with intertissue (between the LMS and normal tissue) variations.47 The findings of their study suggest the relatively homogeneous expression of genes within tumors, although some differences in different parts of some tumors are likely.
In the current study, differences with regard to gene expression between LMS, normal myometrium, and leiomyoma were identified. Genes expressed uniquely in LMS among these samples were identified, as were genes normally expressed in myometrium that were no longer expressed in the leiomyoma samples. Differences with regard to gene expression among different tumors derived from smooth muscle may yield clues to their pathogenesis and potential new treatment approaches. The genes identified herein may be useful in the diagnosis and treatment of LMS as well as in studies of the basic biology of LMS. Analysis of a large number of LMS samples and correlation of biologic phenotypes with gene expression patterns may identify clinically meaningful subsets, as has been reported in cases of large cell lymphoma48 and leukemia.49
The authors thank the staff of Gene Logic Inc., Gaithersburg, MD, for performing the gene expression experiments and D. Trussoni and S. Bowell of the University of Minnesota Tissue Procurement Facility for assistance in collecting and processing tissue samples.