Faculty Bibliography

The histologic diagnosis of sellar masses can be challenging, particularly in rare neoplasms and tumors without definitive biomarkers. Moreover, there is significant inter-observer variability in the histopathological diagnosis of many tumors of the CNS, and some rare tumors risk being misclassified. DNA methylation has recently emerged as a useful diagnostic tool. To illustrate the clinical utility of machine-learning-based DNA methylation classifiers, we report a rare case of primary sellar esthesioneuroblastoma histologically mimicking a non-functioning pituitary adenoma. The patient had multiple recurrences, and the resected specimens had unusual histopathology. A portion of the resected sellar lesion was profiled using clinically validated whole-genome DNA methylation and classification. DNA was extracted from the tissue, hybridized on DNA methylation chips, and analyzed using a clinically validated classifier. DNA methylation profiling of the lesion showed that the tumor classified best with the esthesioneuroblastoma reference cohort. This case highlights the difficulty in diagnosing atypical sellar lesions by standard histopathological methods. However, when phenotypic analyses were nonconclusive, DNA methylation profiling resulted in a change in diagnosis. We discuss the growing role of DNA methylation profiling in the classification and diagnosis of CNS tumors, finding that utilization of DNA methylation studies in cases of atypical presentation or diagnostic uncertainty may improve diagnostic accuracy with therapeutic and prognostic implications.

PURPOSE/OBJECTIVE:Lung adenocarcinoma (LUAD) is a clinically heterogenous disease, which is highlighted by the unpredictable recurrence in low-stage tumors and highly variable responses observed in patients treated with immunotherapies, which cannot be explained by mutational profiles. DNA methylation-based classification and understanding of microenviromental heterogeneity may allow stratification into clinically relevant molecular subtypes of LUADs. EXPERIMENTAL DESIGN/METHODS:We characterize the genome-wide DNA methylation landscape of 88 resected LUAD tumors. Exome sequencing focusing on a panel of cancer-related genes was used to genotype these adenocarcinoma samples. Bioinformatic and statistical tools, the immune cell composition, DNA methylation age (DNAm age), and DNA methylation clustering were used to identify clinically relevant subgroups. RESULTS:Deconvolution of DNA methylation data identified immunologically hot and cold subsets of lung adenocarcinomas. Additionally, concurrent factors were analyzed that could affect the immune microenvironment, such as smoking history, ethnicity, or presence of KRAS or TP53 mutations. When the DNAm age was calculated, a lower DNAm age was correlated with the presence of a set of oncogenic drivers, poor overall survival, and specific immune cell populations. Unsupervised DNA methylation clustering identified 6 molecular subgroups of LUAD tumors with distinct clinical and microenvironmental characteristics. CONCLUSIONS:Our results demonstrate that DNA methylation signatures can stratify lung adenocarcinoma into clinically relevant subtypes, and thus such classification of LUAD at the time of resection may lead to better methods in predicting tumor recurrence and therapy responses.

Chromosomal instability (CIN) is a fundamental property of cancer and a key underlying mechanism of tumorigenesis and malignant progression, and has been documented in a wide variety of cancers, including colorectal carcinoma with mutations in genes such as APC. Recent reports have demonstrated that CIN, driven in part by mutations in genes maintaining overall genomic stability, is found in subsets of adult-type diffusely infiltrating gliomas of all histologic and molecular grades, with resulting elevated overall copy number burden, chromothripsis, and poor clinical outcome. Still, relatively few studies have examined the effect of this process, due in part to the difficulty of routinely measuring CIN clinically. Herein, we review the underlying mechanisms of CIN, the relationship between chromosomal instability and malignancy, the prognostic significance and treatment potential in various cancers, systemic disease, and more specifically, in diffusely infiltrating glioma subtypes. While still in the early stages of discovery compared to other solid tumor types in which CIN is a known driver of malignancy, the presence of CIN as an early factor in gliomas may in part explain the ability of these tumors to develop resistance to standard therapy, while also providing a potential molecular target for future therapies.

We analyzed the contributions of structural variants (SVs) to gliomagenesis across 179 pediatric high-grade gliomas (pHGGs). The most recurrent SVs targeted MYC isoforms and receptor tyrosine kinases (RTKs), including an SV amplifying a MYC enhancer in 12% of diffuse midline gliomas (DMG), indicating an underappreciated role for MYC in pHGG. SV signature analysis revealed that tumors with simple signatures were TP53 wild type (TP53^WT) but showed alterations in TP53 pathway members PPM1D and MDM4. Complex signatures were associated with direct aberrations in TP53, CDKN2A and RB1 early in tumor evolution and with later-occurring extrachromosomal amplicons. All pHGGs exhibited at least one simple-SV signature, but complex-SV signatures were primarily restricted to subsets of H3.3^K27M DMGs and hemispheric pHGGs. Importantly, DMGs with complex-SV signatures were associated with shorter overall survival independent of histone mutation and TP53 status. These data provide insight into the impact of SVs on gliomagenesis and the mechanisms that shape them.

Paired-like homeobox 2b (PHOX2B) is an established immunomarker for peripheral neuroblastoma and autonomic nervous system cells. We aimed to evaluate the utility of PHOX2B immunostaining in central nervous system (CNS) tumors with embryonal morphology. Fifty-one tumors were stained with PHOX2B and submitted for whole slide image analysis: 35 CNS tumors with embryonal morphology (31 CNS embryonal tumors and four gliomas); and 16 peripheral neuroblastomas were included for comparison. Diffuse nuclear immunopositivity was observed in all (16/16) neuroblastomas (primary and metastatic). Among CNS embryonal tumors, focal immunoreactivity for PHOX2B was observed in most (5/7) embryonal tumors with multilayered rosettes (ETMR) and a single high-grade neuroepithelial tumor (HGNET) with PLAGL2 amplification; the remaining 27 CNS tumors were essentially immunonegative (<0.05% positive). Among ETMR, PHOX2B expression was observed in a small overall proportion (0.04%-4.94%) of neoplastic cells but focally reached up to 39% in 1 mm 'hot spot' areas. In the PLAGL2-amplified case, 0.09% of the total neoplastic population was immunoreactive, with 0.53% in the 'hot spot' area. Care should be taken in interpreting PHOX2B immunopositivity in a differential diagnosis that includes metastatic neuroblastoma and CNS tumors; focal or patchy expression should not be considered definitively diagnostic of metastatic peripheral neuroblastoma.

Molecular classification has improved diagnosis and treatment for patients with malignant gliomas. However, classification has relied on individual assays that are both costly and slow, leading to frequent delays in treatment. Here, we propose the use of DNA methylation, as an emerging clinical diagnostic platform, to classify gliomas based on major genomic alterations and provide insight into subtype characteristics. We show that using machine learning models, DNA methylation signatures can accurately predict somatic alterations and show improvement over existing classifiers. The established Unified Diagnostic Pipeline (UniD) we develop is rapid and cost-effective for genomic alterations and gene expression subtypes diagnostic at early clinical phase and improves over individual assays currently in clinical use. The significant relationship between genetic alteration and epigenetic signature indicates broad applicability of our approach to other malignancies.

Image-based analysis as a method for mutation detection can be advantageous in settings when tumor tissue is limited or unavailable for direct testing. Here, we utilize two distinct and complementary machine learning methods of analyzing whole slide images (WSI) for predicting mutated BRAF. In the first method, WSI of melanomas from 256 patients were used to train a deep convolutional neural network (CNN) in order to develop a fully automated model that first selects for tumor-rich areas (Area Under the Curve AUC=0.96) then predicts for mutated BRAF (AUC=0.71). Saliency mapping was performed and revealed that pixels corresponding to nuclei were the most relevant to network learning. In the second method, WSI were analyzed using a pathomics pipeline that first annotates nuclei and then quantifies nuclear features, demonstrating that mutated BRAF nuclei were significantly larger and rounder nuclei compared to BRAF WT nuclei. Lastly, we developed a model that combines clinical information, deep learning, and pathomics that improves the predictive performance for mutated BRAF to AUC=0.89. Not only does this provide additional insights on how BRAF mutations affect tumor structural characteristics, machine learning-based analysis of WSI has the potential to be integrated into higher order models for understanding tumor biology.

Central nervous system (CNS) tumor with BCOR internal tandem duplication (BCOR-ITD) have recently been introduced in the 5th edition of the WHO classification of CNS tumors, however, their molecular makeup and clinical characteristics remain widely enigmatic. This is further complicated by the recent discovery of tumors characterized by gene fusions involving BCOR or its homologue BCORL1. We identified a cohort of 206 BCOR altered CNS tumors via DNA methylation profiling and conducted in-depth molecular and clinical characterization in an international effort. By performing t-SNE clustering analysis we found that BCOR-fusion tumors form a distinct cluster (n=61), adjacent to BCOR-ITD cases (n=145). The identified fusion partners of BCOR(L1) included EP300 (n=20), CREBBP (n=5), and NUTM2HP (n=1). Notably, three cases within the BCOR-ITD cluster harbored a c-terminal intragenic deletion within BCOR. With respect to clinical characteristics gender ratio was balanced in BCOR-fusion cases (m/f, 1.1), whereas predominance of male patients was observed in the BCOR-ITD group (m/f, 1.5). Moreover, age at diagnosis of BCOR-fusion patients was higher as compared to BCOR-ITD cases (15 vs 4.5 years). Interestingly, BCOR-fusion tumors were exclusively found in the supratentorial region being originally diagnosed as ependymomas or gliomas whereas BCOR-ITD emerged across the entire CNS with diverse original diagnoses. 8% of BCOR-ITD and none of BCOR-fusion cases were disseminated at diagnosis. In line with this observation, 40% of first relapses within the BCORITD group were metastatic which was less frequent in BCOR-fusion tumors. Survival estimates demonstrated no differences, generally showing short median PFS (BCOR-fusion, 2 years, n=15; BCOR-ITD, 1.8 years, n=55) and intermediate OS rates (BCOR-fusion, 6.8 years, n=18; BCOR-ITD 6.3 years, n=60). Further molecular and clinical characterization is ongoing potentially revealing first therapeutic leads for these highly aggressive CNS tumor types