The monoclonal origin of the epithelial and mesenchymal compartments has been long established in the literature [6,14]. Prior to the wide availability of whole-genome data, immunohistochemical and mutational analyses demonstrated a high concordance between both compartments [14,15]. There are reports of cases where the above similarities do not hold, which may represent true biclonal “collision tumors” [14,15]. This minority of tumors likely does evolve from two separate progenitor cells, one epithelial and one mesenchymal, that merge to evolve into a single carcinosarcoma tumor. However, this subset represents a small subset of tumors. Furthermore, the prognostic and therapeutic implications of having a monoclonal versus a biclonal tumor remain unclear [16], and thus our review primarily focuses on the instances where a single progenitor cell exists.

More recently, the next-generation sequencing and molecular profiling of carcinosarcomas have provided a better understanding of the molecular signature of gynecologic carcinosarcomas. The most frequently occurring mutations include TP53, PIK3CA, FBXW7, PTEN, KRAS, CCNE1, PPP2R1A, CHD4, and HER2 amplification, among others [17,18]. A summary of the mutational landscape of uterine carcinosarcomas is listed in Table 1. However, the exact percentages of mutations vary across studies, likely reflecting the heterogeneous nature of these tumors [17,19,20,21,22,23,24]. For example, TP53 is consistently the most commonly mutated gene across studies of uterine carcinosarcomas, but the rate varies from 62–91% [25,26]. In uterine carcinosarcomas, the lower end of this range (62%) was found in a study by Gotoh et al. where 35% of the uterine carcinosarcomas included had a low-grade epithelial component [17]. This is in contrast to the higher end of the range (91%), which was found in a study by Cherniack et al., where only 11% of the uterine carcinosarcomas included had a low-grade carcinoma component [20]. This mirrors the TP53 mutation rates found in non-carcinosarcoma endometrial tumors, which are on the order of 10–20% for endometrioid endometrial cancer [27,28], in contrast with mutation rates of approximately 90% in some high-grade endometrial carcinoma histologies [29].

These findings reflect a general trend where the mutational profiles of the carcinosarcoma reflect those of tumors with the corresponding carcinoma component [6,21,30]. Zhao et al. analyzed the mutational profiles of carcinosarcomas with serous and endometrioid carcinoma components and compared them with the profiles of carcinomas with the same histology as the carcinomatous component. They identified eight driver genes with root mutations in their carcinosarcoma samples and analyzed the fraction of tumors with mutations in these genes [21]. They reported that in uterine carcinosarcomas with an endometrioid epithelial part, mutations in PTEN, KRAS, ARID1A, and PIK3CA were prevalent, whereas in uterine carcinosarcomas with a serous epithelial component, mutations in TP53, PIK3CA, FBXW7, CHD4, and PPP2R1A predominated [21]. Other studies of uterine carcinosarcomas containing a serous epithelial component also noted a similar mutational pattern [31,32].

The carcinoma component is probably the defining factor behind the tumor biology and aggressive nature, driving the trans differentiation into a sarcomatous component. Schiff et al. employed comparative genomic hybridization and fluorescence in situ hybridization (FISH) on 30 uterine and ovarian carcinosarcoma samples. They found significant gene amplification, particularly of c-myc within the carcinoma component. They also found a higher proliferation index in the carcinoma component using Ki67 immunohistochemistry. These findings suggest high chromosomal instability and support the more aggressive nature of the carcinoma component relative to the sarcoma component [33]. Moreover, in a recent study, Cuevas et al. were able to induce uterine carcinosarcomas from well-differentiated endometrioid carcinomas in a mouse model through the inactivation of FBXW7 and PTEN in epithelial cells. The genomic analysis of the tumors revealed that most tumors spontaneously acquired a TP53 mutation, suggesting a potential synergistic role of the FBXW7, PTEN/PI3K, and p53 pathways in uterine carcinosarcoma tumorigenesis [34]. Their success in inducing and maintaining a uterine carcinosarcoma by manipulating the epithelial cells further supports the epithelial-driven theory of carcinosarcoma genesis. Two other studies lend further support to the idea that the carcinomatous component is the driver of tumor aggressiveness. A study by Emoto et al. investigated the differences in angiogenesis in the two parts of carcinosarcomas and found a higher VEGF and microvessel density in the epithelial component [35]. Another study likewise reported a higher apoptotic index in the sarcomatous than the carcinomatous component, supporting again that the carcinomatous element plays a key role in the aggressive behavior of this tumor [36]. Clinically, we see evidence of these findings in the fact that most metastatic lesions are carcinomatous in nature [8,14].

In 2013, the Cancer Genome Atlas (TCGA) characterized endometrial carcinomas into four distinct molecular subgroups: POLE ultramutated, microsatellite instability hypermutated, copy number low, and copy number high [39]. Although uterine carcinosarcomas were not included in this analysis, Gotoh et al. subsequently attempted to classify uterine carcinosarcomas using these same molecular characteristics. While many separated into the copy-number-high (serous-like) subgroup, 47% were better classified in one of the three other groups [17]. Similarly, Travaligno et al. confirmed the applicability of the four subgroups in uterine carcinosarcoma, with the copy-number-high group being the predominant subset (91% of those included). Specifically, they found that subgroups with high mutational load (POLE-mutated types and those with high microsatellite instability (MSI-high)) were less common in uterine carcinosarcomas than in other endometrial cancers [25]. The prognostic value of the TCGA classification as applied to uterine carcinosarcoma showed that POLE-mutated carcinosarcomas have an excellent prognosis, similar to that of non-carcinosarcoma endometrial cancers. In contrast, though, the TP53-mutated (copy-number-high/serous-like group) and no specific molecular profile (surrogate of the copy-number-low/endometrioid-like group) groups were associated with a poorer prognosis than their endometrial carcinoma counterparts [40].

Uterine carcinosarcoma incidence rates are five times higher than those of ovarian carcinosarcomas, resulting in fewer studies focusing on ovarian carcinosarcomas in the literature [41,42]. Whereas uterine carcinosarcomas are heterogeneous in their molecular profiling, ovarian carcinosarcomas represent a more homogenous group [7]. In an analysis of the transcriptome of ovarian carcinosarcomas, Gotoh et al. found that the gene expression of ovarian carcinosarcomas most resembled that of high-grade serous ovarian cancer [17]. This is consistent with the fact that the common histology for the carcinoma component is high-grade serous histology [7]. Some studies have shown demonstrated the presence of homologous recombination deficiency in ovarian carcinosarcomas [17,43,44]. Taken together, these findings propose a distinct biological profile for uterine and ovarian carcinosarcomas despite their similar histologic appearance.

Despite the molecular overlap with carcinomas, the mutational profile of carcinosarcomas maintains some key differences that distinguish it from other tumors of the same anatomic site [23]. Unlike most endometrial tumors, the majority of uterine carcinosarcomas contain TP53 and PTEN mutations simultaneously [20,39]. Carcinosarcomas also have been found to have a significantly higher whole-genome doubling than other tumor subtypes, with doubling occurring in 90% of the tumors [20,45]. This percentage is significantly higher than in uterine corpus endometrial carcinomas and ovarian serous carcinomas, which have 22% and 56% whole-genome doubling frequency, respectively [46]. What has really differentiated gynecologic carcinosarcomas from their carcinoma counterparts, however, has been their high rates of mutations in both chromatin-remodeling genes as well as in epithelial–mesenchymal transition (EMT) genes. Jones et al. performed whole-exome sequencing on 22 uterine and ovarian carcinosarcomas and revealed that carcinosarcomas demonstrate one of the highest rates of chromatin remodeling dysregulation of all tumors to date, occurring in approximately two-thirds of cases [22]. ARID1A and ARID1B, key players in the SWI/SNF chromatin remodeling complex, were frequently mutated. Other alterations were also reported, including mutations in histone methyltransferase MLL3, tumor suppressor SPOP, and BAZ1A, a component of the chromatin assembly factor [22]. Mutations in these epigenetic regulators have been shown to be associated with important clinical outcomes in different tumors [47,48]. For instance, mutations in ARID1A and ARID1B have been associated with a decreased survival in patients with neuroblastoma [48]. Their significance in gynecologic carcinosarcomas remains to be elucidated.

Building on this theme, Zhao et al.’s whole-exome sequencing of 68 uterine and ovarian carcinosarcomas confirmed the high mutation rates in histone genes, particularly in genes coding for histone H2A and H2B. They also noted an amplification of the segment of chromosome 6p that contains the histone gene cluster of these genes. When carcinosarcoma cell lines were transfected with mutant H2A and H2B genes, EMT markers showed an accompanying increase, and there was an upregulation in tumor migration and invasion. These findings suggest a potential regulatory role of histones and chromatin remodelers in EMT and sarcomatous transdifferentiation [21].

EMT is a reversible process that involves the transition of a cell from an epithelial to a mesenchymal phenotype. In the context of cancer, EMT plays a critical role in tumor initiation, progression, metastasis, and invasion [49]. This is regulated by a complex system that includes transcriptional factors such as Snail, SLUG, and ZEB1 and ZEB2 and multiple other players such as TGF-B, the JAK/STAT pathway, and microRNAs [50,51]. Particularly, the downregulation of miR-200 family members has been implicated in various aggressive tumors, which is secondary to their role as strong inhibitors of EMT, tumor invasion, and metastasis, among others [51]. The role of EMT in carcinosarcoma has been well-studied, with a special focus on its role in the transdifferentiation of the carcinomatous component into the sarcomatous component. Carcinosarcomas are now widely regarded as one of the best examples of stable EMT [52,53]. Multiple studies have confirmed an association between uterine carcinosarcoma and EMT and show an upregulation in EMT-related genes compared with endometrial carcinomas [20,54,55].

Studies have also confirmed a higher EMT score in the sarcomatous component compared with the epithelial part using both IHC and RT-PCR [32,56,57]. Using transcriptome sequencing, Cherniack et al. found a positive correlation between the EMT score and the presence of heterologous sarcoma histologies as well as a correlation with an increasing proportion of the tumor being made up of a sarcoma component [20]. This study also explored the potential regulators of EMT in uterine carcinosarcoma and reported a key role for the miR-200 family, where its downregulation via promoter hypermethylation is correlated to higher EMT scores and to the presence of a sarcomatous element [20]. Gotoh et al. studied the transcriptome and DNA methylome of carcinosarcomas to confirm the association between EMT and the development of a sarcomatous component. They did not find any correlation between EMT scores and the molecular subtypes of uterine carcinosarcomas previously described [17]. Interestingly, they also did not find an association between EMT and CTNNB1-activating mutations, which have been associated with an induction of EMT in various other tumors [58,59]. Rather, they found CTNNB1 to be associated with the hypomethylation of members of the miR200 family, a finding that was somewhat paradoxical [17]. This likely reflects the complexity of EMT regulators on both a genetic and epigenetic level.

Several other genes have been implicated in EMT in carcinosarcomas. These include HMGA2 as a regulator of Snail expression and the expression of its downstream effectors and ALK as an inducer of EMT and inhibitor of apoptosis [56,60]. It is worth noting that the majority of the molecular and genetic studies investigating EMT in carcinosarcomas have been mostly limited to uterine carcinosarcomas, and very few include ovarian carcinosarcomas. Whether these molecular findings apply to ovarian carcinosarcomas remains to be seen. Lastly, one important corollary to the above discussion is a better understanding of the differences between the carcinomatous and sarcomatous components within the same tumor. Some differences may be due to EMT, but there are likely some other differences between the carcinomatous and sarcomatous components that are present that are independent of EMT. This molecular information may ultimately provide important insights into novel approaches for treatments of this highly aggressive but somewhat unique endometrial carcinoma.