Molecular Genetic Pathology of Soft Tissue Tumors

Small Genetic Variants

Mutations affecting a single nucleotide (SNV) or insertion/deletion ( indel ) of up to 10,000 nucleotides are often referred to as small genetic variants. Such mutations are common in neoplasia as well as in constitutional DNA. Indeed, any individual is thought to deviate from the reference genome at 4 to 5 million sites and to have 2000 to 2500 larger structural variants (gains/losses) affecting up to 20 million nucleotides. Some single nucleotide variants (SNVs) are fairly common (at least 1% of population)—also known as single nucleotide polymorphisms (SNPs)—and may confer increased risk for certain diseases. Neoplastic cells may harbor thousands of SNVs and indels that are not seen in the corresponding constitutional DNA. The vast majority of these mutations are located outside the coding regions, although this does not prevent them from having a significant impact on tumorigenesis; for example, SNVs in the promoter region of TERT increase the affinity for certain transcription factors, a common finding in myxoid liposarcoma and solitary fibrous tumor. Still, the mutations that lead to protein-level changes have attracted the most attention. SNVs could result in the exchange of a single amino acid (“nonsynonymous exonic variants”), introduce a premature stop codon, or abolish a splice recognition site. Similarly, indels could lead to frameshift mutations, truncated proteins, or proteins with novel amino acid sequences. Consequently, SNVs and indels can activate oncogenes, as well inactivate tumor suppressor genes. Only a minority of all the exonic nonsynonymous SNVs and indels found in neoplasms have a major impact on tumor development (driver mutations), while the remaining mutational load represents mutations occurring before neoplastic transformation or noise caused by increased genetic instability (passenger mutations). Several bioinformatic tools can predict the functional outcome of a given mutation, and numerous databases, such as COSMIC ( http://cancer.sanger.ac.uk/cosmic ), list reported mutations by gene and disease, greatly improving the interpretation of detected mutations. However, many mutations can still be difficult to evaluate and may require functional analysis in an experimental system.

Whole exome sequencing efforts initially focused on common epithelial malignancies, such as carcinomas of the breast, colon, and lung. When similar large-scale studies were done on soft tissue sarcomas, some interesting differences were noticed. First, SNVs and indels are much less common in sarcomas than in carcinomas, a difference that could reflect that many carcinomas develop through a gradual transformation of a normal cell, whereas most sarcomas seem to start as malignant lesions, without a recognizable dysplastic or benign precursor lesion. Another possible explanation is that carcinomas develop from stemlike progenitor cells that have undergone numerous cell divisions, each of which may generate SNVs and indels, before transformation, while the turnover of putative mesenchymal progenitor cells is much slower. Second, most sarcomas seem to have few and infrequent recurrent mutations, a phenomenon that may be explained by the presence of other strong driver mutations (see later). However, important exceptions include the frequent KIT and PDGFRA mutations in gastrointestinal stromal tumors (GISTs), RAS signaling pathway mutations in embryonal rhabdomyosarcoma, and MYOD1 mutations in spindle cell rhabdomyosarcoma. Still, there are comparatively few whole exome and whole genome sequencing analyses of sarcomas. Benign soft tissue tumors are even less extensively investigated with regard to somatic SNVs and indels. Frequent and recurrent mutations have thus far been observed in only a handful of benign lesions (e.g., of CTNNB1 in desmoid fibromatosis, NF2 in schwannoma, and PRKD2 in angiolipoma).

Although SNVs and indels presently have a limited impact on soft tissue tumor diagnostics, they may provide important information on treatment response. The prime example is GIST; not only the absence or presence of mutations, but also their precise location in the KIT and PDGFRA genes shows strong correlation with response to treatment with different tyrosine kinase inhibitors. Even if actionable mutations are otherwise rare in sarcomas, targeted treatment based on the presence of certain mutations might improve outcome. Therefore, in patients responding poorly to conventional treatment, mutational screening should be considered.

Chromosomal Imbalances

The term chromosomal imbalance is used here to refer to a structural or numeric rearrangement resulting in a quantitative deviation from the normal diploid state. Such chromosomal imbalances may theoretically range from gain or loss of a single nucleotide to whole chromosomes, but it is reasonable to reserve the term for rearrangements affecting at least an entire gene.

Numeric chromosomal aberrations are poorly tolerated at the organism level but are common in tumors, ranging from gain or loss of individual chromosomes ( aneusomy ) to gain or loss of one or more copies of the entire genome ( ploidy shift ). Chromosome counts as low as 23 and as high as 400 or more have been detected at chromosome banding analysis of tumors ( Fig. 4.1 ). In soft tissue tumors, numeric chromosome aberrations are found in almost two-thirds of all cases subjected to chromosome banding analysis, being much more common among sarcomas than among benign soft tissue tumors, about 90% versus one-third. The difference between benign and malignant soft tissue tumors becomes even more obvious when considering only tumors with chromosome numbers below 45 or above 47; such numbers are seen in only 10% of the benign lesions but in two-thirds of sarcomas. Aneusomies probably arise through incomplete separation of homologues at mitosis, a phenomenon facilitated by a variety of neoplasia-associated disturbances (e.g., of centrioles, centromeres, and telomere status) and tolerated through mutations in various caretaker genes. Gain of one or more complete sets of all chromosomes, polyploidization, is strongly associated with malignancy and is most likely caused by accidental events such as abortive mitoses, cell fusions, and endoreduplication. Also, more extensive, meiosis-like, loss of chromosomes ( near-haploidization ) has been observed in soft tissue tumors, notably inflammatory leiomyosarcoma and undifferentiated pleomorphic sarcoma; the mechanisms behind this phenomenon remain unknown.

The pathogenetic consequences of aneusomies are difficult to assess because they affect hundreds to thousands of genes. In general, however, gene expression levels vary with the number of copies; that is, trisomies and monosomies result in increased and decreased, respectively, expression of many, but not all, genes located on these chromosomes. In some cases the outcome of monosomies can be reduced to represent one step in the biallelic inactivation of one or more tumor suppressor genes. A good example is schwannoma , where inactivation of the NF2 gene, which maps to chromosome band 22q11, could be achieved through any combination of monosomy 22, partial deletions of chromosome arm 22q, and structural variants targeting NF2 specifically. In other tumors, where monosomy and partial deletions alternate, such as spindle cell lipomas, which display either monosomy 13 or partial deletions, with 13q14 as a minimal shared target, the remaining copy of chromosome 13 seems intact, suggesting that loss of one copy (haploinsufficiency) is enough for tumorigenesis. Further, some inactivating mutations, such as those affecting the CDKN2A , CDKN2B , and MTAP loci in chromosome band 9p22, typically do not affect larger segments of chromosome 9, indicating that loss of adjacent segments are deleterious.

Gain of entire chromosomes is even more difficult to reduce to single-gene effects, but it has been suggested that extra copies of chromosome 8 could substitute for gene fusions activating PLAG1 , which maps to chromosome 8, in lipoblastoma . However, the significance of most characteristic numeric aberrations in soft tissue tumors, such as extra copies of chromosome 8 in embryonal rhabdomyosarcoma or trisomies occurring as secondary changes in myxoid liposarcoma or infantile fibrosarcoma, remains poorly understood.

Also, chromosomal imbalances resulting from structural rearrangements are common in soft tissue tumors, especially in sarcomas. For several subtypes, such as GIST, it has even been suggested that the overall number of structural rearrangements leading to imbalances could be used to predict patient outcome. Two types of chromosomal imbalance have attracted particular attention: homozygous deletions and gene amplifications. Homozygous deletions are relatively rare and do not always have to affect bona fide tumor suppressor genes; indeed, some homozygous deletions are constitutional variants without phenotypic consequences. However, pinpointing homozygous deletions in tumors has been instrumental in detecting many of the classic tumor suppressor genes, such as RB1 and SMARCB1. Because men usually have only one X chromosome, a single deletion event on this chromosome will have the same effect as a homozygous deletion on an autosome (i.e., complete loss of one or more genes). Several potential targets for such deletions on the X chromosome have been described in soft tissue tumors, notably the DMD gene in myogenic sarcomas and sclerosing epithelioid fibrosarcoma.

Gene amplification is poorly defined but usually refers to a selective, more than three- to fivefold gain of a DNA sequence relative to adjacent sequences on the same chromosome. Different types of chromosome rearrangement are associated with gene amplification: double minutes (dmin), homogeneously staining regions (hsr), and ring chromosomes, all arising through different mechanisms. The presence of hsr or dmin is strongly associated with a malignant phenotype, whereas ring chromosomes are found in both high-grade malignant and benign/low-grade malignant soft tissue tumors; the latter tumors, however, have a clear propensity for transforming into high-grade tumors, probably because of the intrinsic mitotic instability of ring chromosomes. Ring chromosomes are particularly common among atypical lipomatous tumors/dedifferentiated liposarcomas and dermatofibrosarcoma protuberans, where they occur in 80% and 50% of cases, respectively, and often are seen as the sole cytogenetic aberration.

Another type of chromosomal imbalance, which does not change the gene copy number, is uniparental isodisomy (UPiD), which may affect entire chromosomes or parts of chromosomes. For UPiDs to occur, one normal copy must be lost, followed by duplication of the remaining copy. The effect could be the unmasking of a recessive mutation or, if imprinted genetic loci are involved, deregulated expression of genes inherited from the mother or father. The most commonly affected chromosomal region in soft tissue tumors, as well as in a variety of pediatric malignancies, is 11p. This chromosome arm contains a set of paternally and maternally imprinted loci in band 11p15, including the IGF2 gene, which is expressed exclusively on the copy inherited from the father. Various combinations of loss of the maternal copy of 11p, with or without duplication of the paternal copy, are common in embryonal rhabdomyosarcoma.

Lastly, it should be emphasized that the functional outcome of some imbalances, in particular deletions, does not necessarily need to be the copy number change as such. Each structural rearrangement resulting in loss or gain of chromosomal material could also lead to the juxtapositioning of two genes, one in each breakpoint. Indeed, the HAS2-PLAG1 fusion in lipoblastoma is typically created through an interstitial deletion of the intervening sequences on chromosome 8.

Gene Fusions

Structural chromosome rearrangements reshuffle the genetic material and may thus result in the juxtapositioning of (parts of) two genes. This in turn may lead to the translation of a deregulated and/or chimeric protein ( Fig. 4.2 ). Such gene fusions have been described in all types of neoplasia, including benign as well as malignant lesions. The pathogenetic impact is unknown for most of the gene fusions that have been reported in the literature, the vast majority of which were detected through large-scale deep sequencing studies. Many have only been described once and are accompanied by numerous other chromosome-level mutations and thus likely represent passenger events. Other fusions are repeatedly detected, are often restricted to one or a few morphologic subtypes, and are associated with relatively few other mutations, all suggesting that they constitute strong driver mutations; this is especially true for soft tissue tumors, where characteristic gene fusions abound. The indirect support for an important role in tumorigenesis has been further supported in a few cases by results from in vitro studies and from experimental animal models, showing that the gene fusion, at least if it occurs in a permissive cellular context, is sometimes sufficient for malignant transformation. It is important to emphasize, however, that several pathogenetically important gene fusions are present also in tumors with highly complex genomes, suggesting either that those gene fusions are weak transformers or that they facilitate the occurrence of other mutations. Further, the strong impact of the gene fusions on the tumor cells, coupled with chimeric genes being specific for tumor cells, makes them attractive as potential targets for treatment. Indeed, pharmacologic treatment of sarcomas displaying fusions that activate protein kinases (e.g., ALK) or growth factors (e.g., PDGFB) is already in clinical use.

The first gene fusions to be detected in soft tissue tumors were EWSR1-FLI1 in Ewing sarcoma in 1992 and FUS-DDIT3 in myxoid liposarcomas in 1993. Since then, almost 200 different gene fusions have been found, more than half being recurrent in a specific subtype; about one-third of all soft tissue tumor subtypes display one or more recurrent gene fusions. Table 4.2 shows gene fusions estimated to occur in at least 10% of a specific tumor type.

Gene fusions were first discovered when the molecular outcomes of recurrent cytogenetic aberrations in particular translocations were pursued. At chromosome banding analysis, many of these cytogenetic rearrangements appear balanced; that is, no material seems to be lost or gained, but later analyses at the molecular level show that the recombination causing the fusion is frequently associated with smaller or larger deletions. Often the parts of the genes not included in the fusions are deleted, such as of the FUS and CREB3L2 genes in low-grade fibromyxoid sarcoma. Furthermore, some fusions, such as ASPSCR1 - TFE3 in alveolar soft tissue sarcoma, almost always arise through an unbalanced translocation, possibly adding to the transforming impact by simultaneously creating a fusion gene and deleting or gaining syntenic genes. Other fusions, however, such as COL1A1-PDGFB in dermatofibrosarcoma protuberans and PAX7-FOXO1 in alveolar rhabdomyosarcoma, are typically amplified, either in ring chromosomes or in double minutes. Many sarcoma-associated gene fusions could thus be indirectly detected by genomic arrays, revealing copy number shifts at the sites of fusion.

Chromosome Banding Analysis

Chromosome banding analysis is an excellent screening method for detecting both numeric and structural chromosome aberrations and was for many years the main method to identify new genetic subgroups among soft tissue tumors. It can only be performed on cells in mitosis, more specifically at the metaphase stage, when the chromosomes are contracted enough to be visualized under the microscope. Thus it requires access to fresh tumor tissue, obtained within 2 to 4 days after sampling. After mechanical and enzymatic disaggregation of the sample, cells can be cultured, typically for 1 to 7 days, to achieve metaphase spreads of sufficient quality and quantity.

The band staining of metaphase chromosomes can be achieved through a number of techniques. The most common, G-banding, is obtained through pretreatment with a saline solution or a proteolytic enzyme, followed by staining with Giemsa or similar stains. Several logistical and technical drawbacks hamper the use of chromosome banding analysis in the clinic. First, the need for fresh samples taken under sterile conditions demands efficient transportation from the surgeon or pathologist to the genetic laboratory. Second, even in cytogenetic laboratories with extensive experience, the analysis fails, usually because of overgrowth of normal stromal cells, in 20% to 35% of cases. Third, it is a work-intensive and relatively slow method compared with molecular and molecular cytogenetic techniques. Fourth, the resolution level is poor; structural rearrangements affecting less than 5 to 10 Mb (a chromosome band averages 10 Mb) cannot be detected. Finally, it has proved difficult to obtain tumor-representative karyotypes from fine-needle and core-needle biopsies.

Although chromosome banding techniques are thus increasingly being exchanged for other methods, they are still widely used clinically as well as for scientific purposes and continue to provide important reference data for other genetic studies; more than 2500 soft tissue tumors with abnormal karyotypes have been reported in the literature. In addition, chromosomal banding has shaped the terminology of cancer genetics, emphasizing the importance of being familiar with basic aspects of cytogenetic nomenclature. The autosomal chromosomes in principle are numbered according to size, from 1 to 22. Each chromosome is divided into two arms, separated by the centromere; the shorter, upper arm is called p and the lower, longer arm, q . Each arm is divided into one to four regions, each of which is further subdivided into bands ; regions and bands are numbered from the centromere toward the telomere. Thus, 1p34 denotes chromosome band 4 in region 3 on the short arm of chromosome 1 ( Fig. 4.3A ). Numeric aberrations are specified by a plus or minus sign; for example, +8 and −8 denote gain and loss, respectively, of one copy of chromosome 8 ( Fig. 4.3B ). Structural rearrangements are denoted by an abbreviation for the type of rearrangement ( Table 4.3 ); for example, a t(12;16)(q13;p11) denotes a balanced translocation between chromosomes 12 and 16 with breakpoints in bands q13 and p11, respectively. In a karyotype, which is the sum of all observed clonal changes in a sample, the chromosome number is specified first, followed by the sex chromosome complement, and then by the clonal aberrations observed ( Fig. 4.3B ). Rules for how to report karyotypes and details concerning the nomenclature can be found in the International System for Human Cytogenetic Nomenclature (ISCN, 2016). ISCN also provides directions on how to report results obtained through in situ hybridization, microarrays, and sequence-based assays.

Genomic Arrays

Genomic imbalances (i.e., gains and losses of chromosomal segments) in tumor cells may be detected by hybridizing extracted DNA to defined short DNA fragments, known as probes, attached to a surface, so-called genomic arrays ( Table 4.4 ). The signal intensity depends on the amount of DNA attaching to a certain probe; chromosomal segments that are under- or overrepresented in relation to the average copy number (which depends on the ploidy level of the tumor cells) will thus be recorded as lost or gained. By adding probes that detect SNPs, copy-neutral loss of heterozygosity also can be detected, and the copy number state in aneuploid tumors can be more readily appreciated (see Fig. 4.1 ). The resolution of the analysis depends chiefly on the number and chromosomal distribution of probes, typically amounting to more than 1 million in modern, high-resolution arrays; the information is thus at the exon level in such arrays. Although initially developed for DNA of high quality from fresh or frozen tissue samples, platforms have now been developed that work well with DNA from formalin-fixed, paraffin-embedded (FFPE) samples. Lastly, the amount of input DNA needed for the analysis has kept decreasing, making genomic arrays highly efficient also for analysis of preoperative needle biopsies.

The main conceptual limit of genomic arrays is that they fail to identify balanced chromosomal rearrangements. Thus, most gene fusions associated with soft tissue tumors cannot be detected. However, as previously mentioned, some characteristic fusions are almost always amplified in the tumor cells, whereas other genes involved in fusions often display partial deletions; thus they will then be indirectly identified as copy number shifts in or near the respective genes. Another obstacle is that the organization of the genome (i.e., how different parts of the genome are attached to each other) is not visualized. Thus, coamplified sequences in ring chromosomes in well-differentiated liposarcomas, for example, are seen as separate amplicons in different chromosomes. The major clinical drawback, however, is that the results are highly dependent on the admixture of normal cells to the sample from which DNA was extracted; if tumor cells constitute less than 15% to 20% of the cells in a sample, tumor-associated imbalances will not be detected.

Despite these technical and biologic issues, genomic arrays provide a useful screening method for soft tissue tumors. Unfortunately, comprehensive databases on the copy number profiles of soft tissue tumors are lacking.