Abstract
Background/Aim: Compared to leiomyomas, smooth muscle tumors of uncertain malignant potential (STUMP), and leiomyosarcomas (LMS) originating from the Muellerian duct are very rare. Their molecular pathogenesis remains poorly understood. The present article aims at performing genetic analyses of these tumors that may help assist histopathological examination. Materials and Methods: Ten tumors (four STUMP and six LMS) were investigated by copy number arrays. Results: Two tumors, both classified as STUMP were shown to carry MED12 mutations with one of them presenting with a detectable copy number alteration. All other tumors had multiple copy number changes with a clear predominance of losses. Five chromosomal arms (1p, 13q, 14q, 16q, 22q) were affected by overlapping lost segments in at least four tumors including two cases with biallelic losses of the retinoblastoma gene locus. Conclusion: Besides the general presence of copy number alterations and particular genetic alterations, heterogeneity and ongoing karyotypic evolution indicate malignancy or approaching malignancy.
Recently, research into the patho-biology of uterine smooth muscle tumors has gained considerable interest when the FDA warned against power morcellation of presumed uterine fibroids because of the risk of spreading of unexpected malignant tumors (1-3). This warning has led to several recommendations (4), but there is still a consistent lack of knowledge on the pathogenesis of uterine smooth muscle tumors.
Histologically, among the specific group, benign tumors (leiomyomas and their variants) and clearly malignant lesions (leiomyosarcomas, LMS), as well as a group of tumors with uncertain malignant potential (STUMP) are distinguished. Recent results of molecular investigations suggest that all these tumors constitute pathogenetically heterogeneous groups, subsets of which can be distinguished based on the presence of different driver mutations.
Based on the results of genetic analyses even the group of benign tumors, i.e. leiomyomas, consists of several genetic sub-groups the two largest of which are characterized by either of two mutually exclusive driver mutations i.e. rearrangements of the high-mobility group protein AT-hook 2 gene (HMGA2) locus usually accompanied by cytogenetically visible chromosomal alterations affecting chromosomal bands 12q14~15 (5) and mutations of mediator subcomplex 12 (MED12) (6-8). Other mutations like e.g. rearrangements of HMGA1, a gene closely related to HMGA2, and deletions of the long arm of chromosome 7 may either occur as secondary genomic alterations or as rare driver mutations (7, 9-11). Of note, recent results obtained by whole-genome sequencing as well as by array hybridization showed often highly complex chromosomal alterations that were either much more complex than it had been expected based on classical cytogenetics or even had escaped detection by classical cytogenetics at all (12, 13). To the best of our knowledge, uterine leiomyomas (UL) were the first benign tumors showing a phenomenon known as “firestorms” or “chromothripsis” (7) i.e. complex patterns of segmental chromosomal gains and losses affecting only a few chromosomes or their segments, respectively (14, 15). This phenomenon often co-occurs with HMGA1 or HMGA2 re-arrangements without being exclusively restricted to these UL. It can either affect chromosomal regions typically involved in recurrent clonal chromosomal deviations or affect only regions usually not rearranged in UL.
Besides “ordinary” leiomyomas not otherwise specified (NOS, ICD-O: 8890/0) roughly 10% of the UL are classified as variants mainly based on unusual histological patterns as e.g. cellular (ICD-O: 8892/0) or epithelioid leiomyomas (ICD-O: 8891/0) or those “with bizarre nuclei” (ICD-O: 8893/0). It has been suggested that these histologically-variant leiomyomas often display unorthodox genetic changes that were not described before in UL like extended uniparental disomies (16).
Akin to uterine leiomyomas, clearly malignant lesions, as well, do not seem to constitute a genetically unique entity when investigated by molecular methods. MED12 mutations not distinguishable from those seen in leiomyomas can be found in a considerable percentage of leiomyosarcomas and STUMP as revealed by a couple of independent studies (17-23). These latter cases suggest a possible origin of a subset of LMS from “ordinary” leiomyomas likely as a result of additional genetic alterations occurring in MED12-mutated UL. Accordingly, LMS with areas resembling UL (24, 25) support the assumption that in rare cases a genetic evolution within the tumor cell population leads to focal malignant transformation due to yet unidentified genetic alterations. Accordingly, it seems tempting to speculate that STUMPs carrying MED12 mutations can constitute real “borderline” lesions with the potential to undergo full malignant transformation. There is little doubt, however, that the LMS and STUMP can also arise primarily and independent of pre-existing UL.
In general, the identification of genetic alterations that are likely to accompany and cause malignant and borderline uterine smooth muscle tumors is still in its infancy. Recently, it has been suggested that genomic gain and loss profiles obtained using array-CGH can be used to calculate a so-called genomic index defined as A2/C where A gives the total number of alterations (segmental gains and losses), and C the number of chromosomes involved (26). This index is supposed to split STUMP into a benign group with scarce chromosomal alterations akin to leiomyoma and a malignant group with high chromosomal instability akin to leiomyosarcoma (26). In order to extend and reproduce these latter findings we performed array-based analyses of ten smooth muscle tumors including STUMP and LMS, likely to have originated from the Muellerian duct. The results delineate patterns of genetic alterations that, as an adjunct to histopathology, might serve as hallmarks of malignancy.
Materials and Methods
Tumor samples. Ten smooth muscle tumors of proposed Muellerian duct origin were analyzed. All tumors had been classified as either STUMP or LMS (Table I) according to World Health Organization classification of tumors of female reproductive organs (27). Samples had been formalin-fixed and paraffin-embedded (FFPE) for routine histological diagnosis prior to subsequent molecular investigations performed herein.
Histological examination. For diagnostic purposes tumor samples were fixed in paraformaldehyde (4% in PBS) and processed for paraffin embedding. Tissue sections (1-2 μm) were de-paraffinized in xylene, rehydrated through a series of ethanol, and stained with hematoxylin and eosin (H&E) for histological examination.
DNA isolation and quantification. DNA from the FFPE samples was isolated using the QIAamp DNA Mini Kit (Qiagen, Hilden, Germany) on a QIACube (Qiagen), according to the manufacturer's instructions. The amount of double-stranded DNA was measured using the Qubit dsDNA HS Assay Kit and a Qubit Fluorometer (Life Technologies, Carlsbad, CA, USA).
MED12 mutation analysis. For PCR amplification 1,000 ng of genomic template DNA were used. Primers to amplify the desired human PCR fragment of the MED12 gene were those recently described (6, 8). Subsequently, PCR-products were separated by agarose gel-electrophoresis and the desired DNA fragments/bands were extracted by a QIAquick Gel Extraction Kit (Qiagen) using a QIACube (Qiagen) according to manufacturer's instructions. DNA sequencing of the purified PCR-products was performed by GATC Biotech (Konstanz, Germany).
MIP assay and array hybridization. The OncoScan FFPE Assay (Affymetrix, Santa Clara, CA) is based on Molecular Inversion Probe (MIP) technology and offers the detection of genome-wide copy number and copy-neutral LOH. More than 330,000 MIPs result in a 300-kb genome-wide copy number resolution and an enhanced copy number resolution of 50-100 kb in ~900 cancer genes. Labelling of 80 ng dsDNA and array hybridization was performed following the manufacturer's instructions. After staining and washing using a GeneChip Fluidics Station 450 (Affymetrix) the arrays were scanned by an Affymetrix 3000 7G scanner. Arrays were analyzed through the Nexus Express Software for OncoScan (BioDiscovery, El Segundo, CA, USA).
RNA isolation. Total RNA from frozen tissue samples was isolated using a RNeasy Mini Kit (Qiagen, Hilden, Germany) on a QIACube (Qiagen) according to manufacturer's instructions and DNase I digestion was performed.
cDNA-synthesis. Two hundred and fifty nanograms of total RNA were reverse-transcribed with M-MLV reverse transcriptase (Invitrogen, Karlsruhe, Germany), RNase Out (Invitrogen), random hexamers and dNTPs according to the manufacturer's instructions. RNA was denatured at 65°C for 5 min and subsequently kept on ice for 1 min. After adding the enzyme to the RNA primer mixes, samples were incubated for 10 min at 25°C to allow annealing of the random hexamers. Reverse transcription was performed at 37°C for 50 min followed by inactivation of the reverse transcriptase at 70°C for 15 min.
Expression of HMGA2 mRNA. Relative quantification of transcription levels was carried-out by real-time PCR analyses using the Applied Biosystems 7300 Real-Time PCR system (Applied Biosystems, Darmstadt, Germany). For quantification of HMGA2 mRNA (Hs00171569) a commercially available gene expression assay (Applied Biosystems) was used. HPRT served as endogenous control as described before (28).
Overview of examined samples.
Results
Previous histological examination allowed selecting for tumors classified as STUMP or LMS. A total of ten consecutive malignant or diagnostically challenging tumors of smooth muscle origin were collected. Upon histological examination all these tumors were different from ordinary leiomyomas or their variants, respectively. Based on current WHO classification, four tumors were smooth muscle tumors of uncertain malignant potential (STUMP) and six were leiomyosarcomas of different grades (Table I).
Series included two tumors with MED12 mutations but none showing evidence for HMGA2 rearrangements. All tumors were tested for mutations of mediator sub-complex (MED12) that were found in two of them (UT4 and UT9, Figure 1). Both had a single-base exchange (G>A and G>T, respectively) at position nt.130 representing frequent mutations in ordinary leiomyomas. Furthermore, expression analysis of HMGA2 mRNA was performed by qRT-PCR in all cases to detect rearrangements of HMGA2 leading to its abundant re-expression. Compared to appropriate positive and negative controls none of the tumors presented with an elevated level of HMGA2 transcripts.
Copy number alterations were seen in all but one tumor. All but one tumor (UT4) displayed gains and/or losses detectable by the copy number array analyses and in eight of the tumors more than three imbalances were noted resulting in an “unquiet” signal pattern. For comparison, the genomic overviews of all cases are displayed in Figure 2. All genetic imbalances were simple gains and losses, respectively, whereas no high-level amplifications were observed. Moreover, losses clearly predominated with only a few gains observed. All copy number alterations identified herein were accompanied by corresponding changes of the B-allele frequency (BAF). For none of the gains overlapping segments in more than two cases were observed. In contrast, five chromosomal arms (1p, 13q, 14q, 16q, and 22q) were affected by losses of overlapping chromosomal segments in at least four of the tumors and are herein mentioned in detail. Moreover, there was even one tumor (UT14) whith a hypodiploid karyotype resulting from the apparently complete loss of eleven chromosomes.
The chromosome most often affected by copy number alterations was chromosome 1 showing imbalances in 7/10 tumors (Figure 3A) including two STUMP (UT2, UT28). Whereas in six of these cases exclusively losses were noted, one LMS (UT3) showed gains and losses (Figure 3A). For a segment of the short arm of chromosome 1 (1p32.2-1p34.3) even four copies were noted in a subset of the tumor cell population (Figure 3B). Overlapping segments affected by losses (1p36.22-1pter and 1p13.1-1p13.3) were identified in four cases each. As to the long arm of chromosome 1, the fumarate hydratase gene locus (FH) assigned to 1q43 was affected in two cases only (heterozygous loss in UT2 and UT14).
Other losses with overlapping segments observed in at least four tumors each were noted on chromosomal arms 13q, 14q, 16q, and 22q. As to chromosomes 13, large deletions on the long arm with an overlapping segment assigned to 13q14.13-13q31.2 were observed in four cases (Figure 4A). This commonly deleted segment includes the RB-locus at 13q14.2. Interestingly, in two of these tumors (UT6 and UT15) apparently a second small deletion almost exclusively affecting the RB gene had occurred leading to a bi-allelic partial or full deletion of that gene in part of the tumor cell population (Figure 4B).
Several regions of overlapping losses in at least four tumors were also seen on the long arm of chromosome 14 (Figure 5). Besides four tumors that showed large deletions encompassing almost the entire long arm, three tumors presented with much smaller deletions. Out of these, UT6 carried only a very small deletion of RAD51B, a gene known to be rearranged as a result of the t(12;14)(q14-15;q24) frequently seen in UL.
Four tumors displayed overlapping deletions of the long arm of chromosome 16 (Figure 6). Whereas UT14 is haploid for the entire chromosome 16, UT3 shows a deletion of the distal 36.435 MB. Several losses of segments chromosome 16 were observed in UT21 (Figure 6). The largest of these deletions of 18.17 MB overlapped with a 1.11 MB deletion of part of chromosomal band 16q21.1 in UT6.
Four tumors, i.e. UT10, UT14, UT21, and UT28 showed deletions of large parts of the long arm of chromosome 22 with large overlapping regions in 22q12.1-22q12.3 and 22q13.1-22q13.2.
Sub-clonal genetic heterogeneity seen in the majority of the tumors. Based on the assumption that the occurrence and amount of genetic heterogeneity among tumor cells can be a hallmark of intratumoral evolution, we used two criteria to evaluate heterogeneity. Firstly, the CNV arrays used for the present study allow for distinguishing between different frequencies of alterations and different genetic sub-populations, respectively, within the tumor sample under investigation. As shown in Figure 2, many of the tumors were characterized by aberrations showing different “distances from zero” as also reflected by corresponding changes in the BAF lane. Secondly, often gradiental losses at the border of deleted fragments are seen. These gradiental losses can be detected easily by the diagonal placement of the point cloud at one or either sides of deleted segments. In the latter case, these gradients are displayed as “bubbles” in the corresponding BAF panel. Examples are shown in Figure 7A and B.
Within the present series the number of clearly distinguishable sub-populations varied strongly. Whereas no evidence for different genetic sub-populations was obtained in only two of the tumors which were both classified as STUMP (UT2, UT4), all other samples including the two remaining STUMP revealed heterogeneity within the tumor cell population ranging between two and four clearly distinguishable sub-populations. Except for the two STUMP displaying no or only little genetic heterogeneity, six of the remaining tumors showed gradiental losses likely representing an ongoing karyotypic instability.
MED12 mutations in two of the smooth muscle tumors investigated by array hybridization. Both had a single base exchange (UT4: G>A and UT9: G>T) at position nt.130.
B-allele frequency reveals an area of acquired loss of heterozygosity coinciding with normal copy number. As a rule, alterations of the BAF panel coincided with copy number alterations and only those copy number alterations accompanied by corresponding changes of the BAF panel were considered to be of relevance. Nevertheless, UPD in the absence of copy number alterations recently has been reported in an uterine smooth muscle tumor that was classified as an epithelioid variant of a leiomyoma (16). Within the present series one tumor revealed acquired uniparental disomy as well, though it affected a much smaller segment of the genome. Chromosomal region 1p36.13-p34.3 of case UT3 showed a normal copy number but clearly revealed uniparental disomy. As revealed by the position of the BAF lanes, this UPD did not characterize all cell of the sample indicating that it was acquired during tumorigenesis and did not occur in normal cells (Figure 3B).
Discussion
In rare cases of smooth muscle tumors of Muellerian duct origin the correct histological diagnosis remains a diagnostic challenge. In certain cases even a continuum between leiomyomas and leiomyosarcomas seems to exist as histologically witnessed e.g. by areas displaying different degrees of histological abnormalities ranging between benign and malignant areas (24, 25). From a genetic point of view virtually all uterine smooth muscle tumors, benign and malignant, are likely to be characterized by genetic abnormalities distinguishing them from their tissue of origin. Some of these abnormalities act as driver mutations the presence of which often seems to indicate independent pathogenetic entities. In the present study, we used CNV-arrays to detect genomic alterations of malignant and borderline smooth muscle tumors including STUMPs as well as leiomyosarcomas. The data together with those from earlier reported studies on LMS and UL using genomic arrays (13, 16, 17, 26), reveal that even the presence of gross genetic imbalances does not unequivocally indicate malignancy according to the WHO criteria recently proposed (Figure 2 and Table I). Likewise, the genomic index, as suggested by Croce et al. (2015) can be of some help. Nevertheless, in the absence of detectable genomic alterations no such genomic index can be calculated and we feel that due to chromothripsis seen in some “ordinary” UL high genomic indices do not identify malignant lesions with sufficient specificity. On the other hand, some of the mutations primarily found in UL are not restricted to benign growth but occur in subsets of STUMPs and leiomyosarcomas as well. A well-investigated example are mutations of the gene encoding mediator subcomplex 12 (MED12) as initially described in leiomyomas (6). In several subsequent studies their presence in STUMP and LMS has been documented and in the present study as well two STUMP presented with MED12 mutations (17-23). In contrast, to the best of our knowledge, no STUMP or LMS with HMGA2 rearrangement has been reported so far. Another abnormality that was identified both in the study by Croce et al. (26) as well as in our study in LMS but also occurs in UL is the deletion of 22q.
Genomic overview showing gains and losses across the ten smooth muscle tumors investigated. For each of the tumors, the top panel displays the copy number probe intensity calls and the second panel displays the calculations of B-allele frequencies (BAF).
Detection of gains and losses and allele frequencies of chromosome 1 in seven uterine smooth muscle tumors (for details on Table I). (A) Ideogram of chromosome 1 (top) and panels displaying copy number data across chromosome 1 from tumor seven tumor samples. The panel shows the copy number probe intensity calls and the segments affected by gains (blue) and losses (red). (B) Copy number and SNP data across chromosome 1 for tumor UT3. Top: Ideogram of chromosome 1, upper panel: segments affected by copy number alterations with gains shown in blue and losses shown in red. Lower panel: BAF (B-allele frequency) based on calculated SNP allele ratios. Segmental uniparental disomy is noted on part of the short arm of the chromosome which, in a segment of roughly 20 Mb, does not correspond to copy number alterations.
Homozygous deletions of the long arm of chromosome 13 affecting the retinoblastoma gene (RB) locus in two uterine smooth muscle tumors (for details see Table I).
Accordingly, it seems tempting to speculate that some of the genetic abnormalities seen in uterine smooth muscle tumors can either determine benign or malignant growth while others can act as driver mutations of initially benign growth with a certain, albeit low, probability of secondary malignant transformation likely to depend on the occurrence of additional genetic abnormalities. Of note, the example of 22q deletions suggests that these abnormalities of “mixed potential” do not seem to be restricted to MED12 mutations but also include other changes of the tumor genome. Therefore, it currently seems difficult to calculate the “true” frequency of LMS that have originated from pre-existing UL.
Detection of gains and losses of chromosome 14 in seven uterine smooth muscle tumors (for details see Table I).
Detection of gains and losses of chromosome 16 in four uterine smooth muscle tumors (for details see Table I).
Currently, the FDA has addressed the procedure of power morcellation of uterine tumors because of the risk of intraperitoneal spreading of unexpected malignant tumors (1, 29). To avoid iatrogenic spread resulting from morcellation during myomectomy or hysterectomy FDA has warned against the procedure. Simultaneously, additional diagnostic procedures are discussed as e.g. intraoperative biopsies. Given the histological heterogeneity of leiomyosarcomas often displaying only focally a “malignant areas” (24, 25), such a procedure seems less well-suited to distinguish between tumors suited for morcellation or not. The same holds true for genetic criteria as an adjunct to morphological criteria because genetic heterogeneity seems to be a hallmark of malignant transformation in these tumors, too. In addition, the identification of genetic alterations specifically associated with malignancy remains a challenge.
Examples of gradients of the BAF panel: A: UT9, chromosome 7. B: UT21, chromosome 6. A/B, top: ideograms of chromosomes 7 and 6, respectively, with red bars indicating losses and blue bars indicating gains. Ideograms followed by results of copy number probe intensity calls and calculations of B-allele frequencies (BAF).
Nevertheless, as to surgically-removed samples the results of the present study confirm the usefulness of copy number arrays as an adjunct to classical histopathology in this type of tumors. Though, as discussed earlier, the mere number of gains and losses and the resulting “unquietness” of the pattern does not seem to be useful as a stand-alone parameter to diagnose malignancy, certain patterns and markers emerge, that can be used additively to enhance specificity.
As to certain losses, those of parts or the entire long arm of chromosome 14 were a frequent finding. Monosomy 14 has been repeatedly detected in non-uterine leiomyosarcomas and gastrointestinal stromal tumors (GIST) by classical as well as by molecular cytogenetics including cases with monosomy 14 as the only detectable abnormality (30–34). Interestingly, another frequent abnormality identified in the smooth muscle tumors investigated herein, i.e. deletions of parts of the long arm of chromosome 22 is also shared by GIST (31-33). In this latter entity, the occurrence of 22q deletions has been correlated with the transition to an unfavorable cytogenetic sub-pathway (32). On the other hand, loss of 22q has been identified as a recurrent aberration in non-variant uterine leiomyomas as well (35, 36). As a secondary abnormality seen in a subset of the tumor cells monosomy 22 was thought to reflect a preferred pathway in the karyotypic evolution of UL (36). Terminal deletion of 22q was also seen in five cases of “benign metastasizing leiomyoma” (37). Recently, deletions of 22q have been identified as a secondary abnormality in UL by Mehine et al. As the target gene, DEPDC5 has been suggested to act as a tumor suppressor gene the loss of which played a role in the progression of uterine leiomyomas (38). Interestingly, in the same study bi-allelic losses of DEPDC5 were found as well.
Deletions affecting the long arm of chromosome 13 were also frequently observed in the present series of smooth muscle tumors. Of note, deletions of the long arm of chromosome 13, as the sole cytogenetically detectable abnormality, have been reported in a few cases of UL as well (39). In two of the cases carrying 13q deletions investigated herein bi-allelic losses resulting from one large and one small deletion confined to the RB gene or part of it, respectively, have been detected. In both cases the two deletions occurred under a different percentage within the tumor cell population likely to consider them as subsequent steps during tumor progression. A substantial excess risk for the development of uterine leiomyosarcoma has been noted in female retinoblastoma survivors (40). Thus, the RB gene can be considered to be at least one of the targets of 13q deletions in smooth muscle tumors of Muellerian duct origin. Since mutations of the RB locus efficiently resulted in the formation of sarcomas in nude mice (41) it seems reasonable to check tumors with detectable larger deletions of 13q for hidden mutations of the remaining allele in future studies.
Moreover, the near haploid karyotype identified in UT14 deserves particular interest. Recently, we described an epithelioid UL, that showed normal copy numbers for the majority of the chromosomes which, however, displayed uniparental disomy (16). We concluded that in this case a near-haploid chromosomal set, during tumor development, has preceded the duplication of the genome. In contrast, in UT14 a persistently haploid genome without subsequent duplication is noted. The uniparental disomy for part of the short arm of chromosome 1 as observed in case UT3 was likely to have arisen as an acquired abnormality during tumorigenesis by somatic recombination. In general, acquired UPD is not a rare phenomenon in cancer and e.g. has been well-investigated in acute myeloid leukemias where it can affect a variety of chromosomal segments (42-44). Acquired UPD has often been discussed as a driver force in cancer development through different mechanisms [for review see (45)].
In summary, a weakness of the present study is the small number of cases tested. Nevertheless, its results allow for identifying genetic changes that are associated with malignant growth and can be used as additional diagnostic criteria in challenging smooth muscle tumors of Muellerian duct origin. Out of these, extended genetic heterogeneity, as revealed by a different frequency of sub-populations and by gradiential losses as well as deletions of 14q per se may indicate malignancy. On the other hand, the occurrence of genetic alterations shared between UL on the one side and STUMP/LMS on the other side might suggest that LMS occur more often as secondary malignancies than previously estimated and deduced just from those cases carrying MED12 mutations. Among those alterations deletions of the long arm of chromosome 22 seem to exist. As to the risks associated with power morcellation, the number of tumors should be considered as an independent risk factor for malignant disease.
Acknowledgements
This study was supported by Zahnarztpraxis Weiß und Partner, Bremen, and the Bremer Krebsgesellschaft e.V. We also wish to thank Frauke Meyer for her excellent technical assistance.
- Received August 14, 2015.
- Revision received September 10, 2015.
- Accepted September 14, 2015.
- Copyright© 2015 International Institute of Anticancer Research (Dr. John G. Delinassios), All rights reserved
References
In this issue
Jump to section
Related Articles
Cited By...
- Reasons to Reconsider Risk Associated With Power Morcellation of Uterine Fibroids
- Case Report: A Low-grade Uterine Leiomyosarcoma Showing Multiple Genetic Aberrations Including a Bi-allelic Loss of the Retinoblastoma Gene Locus, as well as Germ-line Uniparental Disomy for Part of the Long Arm of Chromosome 22