Intimal sarcomas and undifferentiated cardiac sarcomas carry mutually exclusive MDM2, MDM4, and CDK6 amplifications and share a common DNA methylation signature

Undifferentiated mesenchymal tumors arising from the inner lining (intima) of large arteries are classified as intimal sarcomas (ISA) with MDM2 amplification as their molecular hallmark. Interestingly, undifferentiated pleomorphic sarcomas (UPS) of the heart have recently been suggested to represent the cardiac analog of ISA due to morphological overlap and high prevalence of MDM2 amplifications in both neoplasms. However, little is known about ISAs and cardiac UPS without MDM2 amplifications and molecular data supporting their common classification is sparse. Here, we report a series of 35 cases comprising 25 ISAs of the pulmonary artery, one ISA of the renal artery and 9 UPS of the left atrium. Tumors were analyzed utilizing the Illumina Infinium MethylationEPIC BeadChip array, enabling copy number profile generation and unsupervised DNA methylation analysis. DNA methylation patterns were investigated using t-distributed stochastic neighbor embedding (t-SNE) analysis. Histologically, all ISAs and UPS of the left atrium resembled extra-cardiac UPS. All cases exhibited highly complex karyotypes with overlapping patterns between ISA and UPS. 29/35 cases showed mutually exclusive amplifications in the cell-cycle associated oncogenes MDM2 (25/35), MDM4 (2/35), and CDK6 (2/35). We further observed recurrent co-amplifications in PDGFRA (21/35), CDK4 (15/35), TERT (11/35), HDAC9 (9/35), and CCND1 (4/35). Sporadic co-amplifications occurred in MYC, MYCN, and MET (each 1/35). The tumor suppressor CDKN2A/B was frequently deleted (10/35). Interestingly, DNA methylation profiling (t-SNE) revealed an overlap of ISA and cardiac UPS. This “ISA” methylation signature was distinct from potential histologic and molecular mimics. In conclusion, our data reveal MDM4 and CDK6 amplifications in ISAs and UPS of the left atrium, lacking MDM2 amplification. We further report novel co-amplifications of various oncogenes, which may have therapeutic implications. Finally, the genetic and epigenetic concordance of ISAs and UPS of the left atrium further supports a shared pathogenesis and common classification.


INTRODUCTION
Intimal sarcoma (ISA) is an exceedingly rare undifferentiated sarcoma that arises in the pulmonary artery and less frequently in the aorta or its branches [1]. ISAs behave highly aggressive with a mean patients' survival ranging from 5 to 18 months [1]. Patients with ISA are mostly of middle age at diagnosis and typically present with non-specific symptoms, which sometimes masquerades as thromboembolic disease. Thus, patients are often diagnosed in an advanced disease stage. Furthermore, ISAs are often reported to be resistant to conventional chemotherapy [2].
First recognized by Mandelstamm as pulmonary artery sarcoma from an autopsy in 1923, case reports and small case series have been published for the following decades [3]. The term "intimal" was coined with reference to the attributes that these sarcomas arise from the subendothelial space of arteries, form polypoidal and endoluminal protrusions and spread laterally along the intima of large arteries before they infiltrate beyond the adventitia. However, it has also been noted that sarcomas other than ISA may occasionally exhibit such "intimal" growth pattern [4,5].
Molecular studies on pulmonary artery ISAs revealed a high frequency of MDM2 amplifications, accompanied by coamplifications of CDK4 and PDGFRA [8][9][10][11][12][13][14][15]. In 2014, Neuville et al. discovered MDM2 amplifications in a large proportion of UPS of the heart. Interestingly, many of the MDM2 amplified cardiac UPS presented with histologic features resembling ISA [13]. The authors proposed the concept that these cardiac UPS may represent ISA [16]. However, concerns rose with regards to the non-specific occurrence of MDM2 amplifications in a broad range of tumor types [10]. Moreover, genome-wide copy number analysis has been performed only in a few pulmonary artery ISAs, whereas ISAs of more uncommon sites such as the aorta have not been analyzed to date (Table 1). It was concluded that reclassifying these cardiac UPS as ISA could be premature, besides that fact that the term "intima" would be inappropriate for these tumors from an anatomic point of view [17,18]. Thus, the 4 th edition of the WHO classification of tumors of the heart differentiates the multiple subtypes of cardiac UPS by histopathology, but mentions ISA as a synonym or alternative designation [19,20]. Thus, the relationship between arterial and cardiac ISAs remains incompletely understood.
High density DNA methylation arrays provide a powerful tool for robust molecular tumor classification [21,22]. DNA methylation profiling in sarcomas has defined subtype-specific sarcoma signatures, even within seemingly morphological homogenous entities that would otherwise evade a definite histologic diagnosis. Likewise, DNA methylation profiling has shown morphological heterogeneous tumors to constitute a single molecular subtype [23][24][25][26][27][28][29]. Furthermore, data of these high-density DNA methylation arrays allow genome-wide mapping of copy number variations.
Herein, we comprehensively characterize a cohort of 26 ISAs and 9 UPS of the left atrium by genome-wide copy number analysis and DNA methylation profiling. We sought to further define their molecular alterations and determine whether they share a DNA methylation signature that segregates them from potential histologic mimics.

MATERIAL AND METHODS Sample selection
We collected a cohort of 26 ISAs and 9 UPS of the left atrium from different patients, containing 10 previously published cases [8,9]. Samples were retrieved from the Institute of Pathology of the University Zürich  Table 1.
Diagnoses were established according to the guidelines of the WHO classification for soft tissue and bone tumors (5 th edition) and for tumors of the lung, pleura, thymus and heart (4 th edition) [1,19]. Accordingly, none of the study cases showed a definable line of differentiation. The study was performed in concordance with the guidelines set forth by the local ethics committee of the University of Heidelberg and in accordance with the Declaration of Helsinki.

DNA extraction and quantification
DNA of all tumors was extracted from formalin-fixed paraffin-embedded (FFPE) tissue samples. All tumors included in this study had sufficient tumor material available to prevent extraction of neighboring benign tissue. Areas with highest available tumor content (≥70%) were chosen for extraction of DNA. The Maxwell® 16 FFPE Plus LEV DNA Kit was applied on the automated Maxwell device (Promega, Madison, WI, USA) according to the manufacturer's instructions. Extracted DNA was quantified using the QuantiFast SYBR Green PCR Kit (Qiagen, Duesseldorf, NW, Germany). A minimum of 100 ng DNA was extracted in every case and provided for subsequent array-based DNA methylation analysis.
Genome-wide DNA methylation data generation and preprocessing The total DNA input suitable for DNA methylation profiling ranges from a minimum of 10 ng to 500 ng [30]. In this study, all 35 samples reached a total DNA input of ≥100 ng and therefore were subjected to the Illumina Infinium MethylationEPIC BeadChip array (Illumina, San Diego, USA) analysis at the Genomics and Proteomics Core Facility of the German Cancer Research Center (DKFZ) Heidelberg. To exclude low-quality samples from the cohort, the on-chip quality metrics of all samples were checked and compared to a set of 7,500 pairs of IDAT-files [22]. All 35 samples passed this quality control check. DNA methylation data were normalized by performing background correction and dye bias correction (shifting of negative control probe mean intensity to zero and scaling of normalization control probe mean intensity to 20000, respectively). Probes targeting sex chromosomes, probes containing multiple single nucleotide polymorphisms and those that could not be uniquely mapped were removed. Human reference genome (hg19) was used for the analysis of multi-site mapping. Probes from the EPIC array were excluded if the predecessor Illumina Infinium 450k BeadChip did not cover them, thereby making data generated by both 450k and EPIC feasible for subsequent analyses. In total, 438370 probes were kept for analysis.

Copy number analysis
Copy number plots were generated on methylation array data using the R package 'conumee' after additional baseline correction (https://github. com/dstichel/conumee). Copy number variants were identified by manual C. Koelsche et al.
inspection as previously described [31]. Thresholds for the identification of amplifications and homozygous deletions were derived from the difference of the baseline. Gains/amplifications usually are above a log2 value of 0.4 and losses/deletions usually are below a log2 value of 0.4. Low tumor cell content or subclonal alterations may reduce the amplitude deviation.

Fluorescence In Situ Hybridization
Additional FISH assays were performed for assessing gene copy alterations. FISH on interphase nuclei from FFPE 4 μm sections was performed using custom-designed probes of bacterial artificial chromosomes flanking the target genes CDK6 (7q21.2) and MDM4 (1q32.1). An amplification was defined as the presence of >10 signals (ratio to control probe >10) or tight clustered signals characteristic of homogeneous staining regions.

Study cohort and histopathology
The study included 25 cases with ISAs arising in the pulmonary artery and one case with an ISA of the renal artery. The sex ratio was balanced (male-female ratio, 13:13). Patient's age at presentation ranged from 30 to 83 years, with a median age of 58 years. Furthermore, the study included 5 female and 4 male patients with UPS of the left atrium, among them one young patient with a metastatic bone lesion in the humerus, which was analyzed here. Their age at presentation ranged from 18 to 67 years, with a median age of 38 years. The patient's characteristics are summarized in Table 2.
Morphologically, all 35 cases, some at least focally, showed ISA features characterized by endoluminal growth, fibrin layering with tumor overgrowth and intimal spread (Fig. 1). Overall, the cellularity was variable within tumors. The predominant tumor architecture was loose and pattern-less, although storiform areas and collagenized stroma were recognizable in some cases. The tumor cells appeared mostly spindle shaped (Fig. 2a) and rarely epithelioid (Fig. 2b). Nuclear pleomorphism was evident in all cases. Some case showed prominent bleeding residues and two cases presented with dystrophic calcifications (Fig. 2c). Three cases exhibited patterns reminiscent of distinct soft tissue sarcoma subtypes other than undifferentiated (pleomorphic) sarcoma. One case (ID 141642) presented with myxoid areas with low cellularity, but tumor cell condensation around vessels, resembling myxofibrosarcoma (Fig. 2d). One case (ID 141634) showed Table 2. Patient characteristics and histologic features.

Variables
Artery (n = 26) Heart (n = 9) pseudoangiomatous spaces filled with blood and surrounded by tumor cells, a pattern that to some extend resembled an angiomatoid fibrous histiocytoma (Fig. 2e). Another case (ID 129604) showed prominent plasma cell aggregates and scattered eosinophils within an otherwise hypocellular tumor stroma, resembling inflammatory myofibroblastic tumor (Fig. 2f).

Recurrent gene amplifications beyond MDM2 and PDGFRα
We next analyzed the copy number profiles for amplifications and deletions (Fig. 3). Common DNA methylation signature in intimal sarcomas and undifferentiated pleomorphic sarcomas of the left atrium We next analyzed DNA methylation profiles by t-SNE analysis (Fig. 6). ISAs and UPS of the left atrium formed a unique methylation cluster when compared to potential histopathologic and molecular mimics, thereby demonstrating a distinct "ISA" methylation signature. Interestingly, the sample of a bone metastasis of an ISA overlaid with primary ISA samples, which illustrates the stability of this "ISA" methylation signature. Finally, it is important to note that the aforementioned mimics also constituted subtype specific clusters in the t-SNE plot.

DISCUSSION
In this study we assessed the molecular profiles of ISAs and UPS of the left atrium using genome-wide copy number profiling and unsupervised DNA methylation analysis. ISAs and UPS of the left atrium showed highly complex karyotypes. Our analysis revealed MDM4 and CDK6 amplifications in ISAs and cardiac UPS lacking MDM2 amplifications. Furthermore, their epigenetic patterns were highly overlapping. Thus, our data argue in favor for the proposed concept by Neuville and colleagues that ISAs and UPS of the left atrium may constitute a common entity [13,16]. Previous studies on ISAs and cardiac sarcomas have used array-CGH analysis only in a small number of samples and tested their findings on larger cohorts by fluorescence in situ hybridization analysis. These studies revealed complex karyotypes with highly recurrent amplifications of MDM2 [8][9][10][11][12][13][14][15]. With this approach, however, there is a risk that less frequent copy number variations can be missed. In our study, all 35 cases were tested for genomewide copy number variations. We could confirm MDM2 amplifications in the vast majority of ISAs and additionally identified previously unreported, mutually exclusive MDM4 and CDK6 amplifications in ISAs and UPS of the left atrium. MDM4 and CDK6 immunohistochemistry were strongly positive in cases with such amplifications (data not shown). However, we also noticed a marked MDM4 and CDK6 expression in cases with balanced MDM4 and CDK6 locus. Therefore, we consider MDM4/CDK6 immunohistochemistry not suitable as surrogate for detecting MDM4/CDK6 amplifications.
MDM2 and CDK6 are critical components in the p53 and RB signaling pathway, respectively. These pathways are frequently disrupted in sarcomas with complex karyotypes, probably because subunits of those pathways are also involved in maintaining chromosome integrity [32]. CDK6 and CDK4 form a complex that suppresses RB1, a key element in the RB pathway. Notably, the coamplification of CCND1 and the recurrent deletion of CDKN2A/B also contribute to the disruption of the RB pathway [33]. MDM4 is best characterized for repressing p53 transactivation activity and p53 translational regulation in normal cells under stress [34]. Because of the high frequency of amplifications in these cell cycle pathways, they appear to contribute to the pathogenesis in ISA and UPS of the left atrium.
We also observed highly recurrent amplifications of PDGFRA, as previously reported [12]. Other RTKs were only rarely amplified, e.g., MET. Copy number alterations were rarely observed in the oncogenes MYC and MYCN. Furthermore, we observed frequent TERT amplifications, which confer unlimited proliferation potential to cancer cells through telomere length maintenance. TERT amplifications have been shown to confer the highest telomerase activity among tumors [35]. Finally, we detected HDAC9 amplifications in ISAs. Higher expression of HDAC9 has been associated with poor prognosis in different cancer types [36]. These oncogene amplifications in ISAs and UPS of the left atrium may qualify patients for targeted therapies [37,38]. However, it remains to be determined whether targeted therapies can lead to a treatment response in ISAs [2].
Some cases lacking MDM2, MDM4, or CDK6 amplifications remain ambiguous by copy number profiling. It is conceivable that these cases harbor alternative driver mutations, although ISAs show a relative low mutation rate overall [15]. In our study, cases lacking MDM2, MDM4, or CDK6 amplifications often carried PDGFRA amplifications. However, PDGFRA amplifications are among the most common genetic alterations in cancer. Accordingly, we observed PDGFRA amplifications also in 5/12 leiomyosarcomas (data not shown). Hence, the detection of PDGFRA amplifications alone is non-specific for diagnosing ISA [39]. To some extent, the same concerns apply for MDM2/4 amplifications. ISAs of uncommon sites such as the retroperitoneum, where dedifferentiated liposarcoma is always a strong consideration, are prone to be misdiagnosed, especially in conjunction with an underlying MDM2 amplification [40][41][42]. Thus, clinical correlation is paramount for the diagnosis of ISAs and the isolated use of established molecular markers alone may not be helpful in this context.
We provide evidence that DNA methylation profiling may be valuable in increasing the diagnostic accuracy of these rare tumors. Our methylation analysis led to the identification of a common methylation fingerprint in ISAs and cardiac UPS. We assume this "ISA" methylation signature to be specific. Accordingly, relevant differential diagnoses, e.g., dedifferentiated liposarcomas and extra-cardiac UPS showed distinct methylation signatures compared to ISA [43,44]. Notably, these methylation signatures remain stable over the timeline of the disease, as previously reported in other tumor entities [45]. We could detect the ISA methylation signature even in the bone metastasis of a cardiac UPS. In view of the specificity and stability of this "ISA" DNA methylation signature, DNA methylation profiling may be a valuable ancillary biomarker in ISAs and UPS of the left atrium, especially in cases lacking MDM2, MDM4 or CDK6 amplifications or occurring at unusual sites.