A DNA methylation–based classifier facilitates the differentiation of primary lung and metastatic head and neck cancer with diagnostic accuracy.

The diagnosis of sinonasal tumors is challenging due to a heterogeneous spectrum of various differential diagnoses as well as poorly defined, disputed entities such as sinonasal undifferentiated carcinomas (SNUCs). In this study, we apply a machine learning algorithm based on DNA methylation patterns to classify sinonasal tumors with clinical-grade reliability. We further show that sinonasal tumors with SNUC morphology are not as undifferentiated as their current terminology suggests but rather reassigned to four distinct molecular classes defined by epigenetic, mutational and proteomic profiles. This includes two classes with neuroendocrine differentiation, characterized by IDH2 or SMARCA4/ARID1A mutations with an overall favorable clinical course, one class composed of highly aggressive SMARCB1-deficient carcinomas and another class with tumors that represent potentially previously misclassified adenoid cystic carcinomas. Our findings can aid in improving the diagnostic classification of sinonasal tumors and could help to change the current perception of SNUCs.

Soft-tissue sarcomas (STS) are rare malignancies that account for 1% of adult cancers and comprise more than 50 entities. Current therapeutic options for advanced-stage STS are limited. Immune checkpoint inhibitors targeting the PD-1/PD-L1 signaling axis are being explored as new treatment modality in STS; however, the determinants of response to these agents are largely unknown. Using the sarcoma data set of The Cancer Genome Altas (TCGA) and an independent cohort of untreated high-grade STS, we analyzed DNA copy number status and mRNA expression of PD-L1 in a total of 335 STS cases. Copy number gains (CNG) were detected in 54 TCGA cases (21.1%), of which 21 (8.2%) harbored focal PD-L1 CNG and that were most prevalent in myxofibrosarcoma (35%) and undifferentiated pleomorphic sarcoma (34%). In the untreated high-grade STS cohort, we detected CNG in six cases (7.6%). Analysis of co-amplified genes identified a 5.6-Mb core region comprising 27 genes, including JAK2. Patients with PD-L1 CNG had higher PD-L1 expression compared with STS without CNG (fold change, 1.8; p = 0.02), an effect that was most pronounced in the setting of focal PD-L1 CNG (fold change, 3.0; p = 0.0027). STS with PD-L1 CNG showed a significantly higher mutational load compared with tumors with a diploid PD-L1 locus (median number of mutated genes; 58 vs. 40; p = 3.6E-06), and PD-L1 CNG were associated with inferior survival (HR = 1.82; p = 0.025). In contrast, T-cell infiltrates quantified by mRNA expression of CD3Z were associated with improved survival (HR = 0.88; p = 0.024) and consequently influenced the prognostic power of PD-L1 CNG, with low CD3Z levels conferring poor survival in cases with PD-L1 CNG (HR = 1.8; p = 0.049). These data demonstrate that PD-L1 GNG and elevated expression of PD-L1 occur in a substantial proportion of STS, have prognostic impact that is modulated by T-cell infiltrates, and thus warrant investigation as response predictors for immune checkpoint inhibition.