Extended cold lung preservation with improved mitochondrial health: pre- clinical studies and translation to human lung transplantation (tr)

Chordomas are rare malignant bone sarcomas of the skull-base and spine with significant variability in patient survival that cannot be reliably predicted using clinical factors or genomic alterations. Here, we show, for the first time, two distinct epigenetic chordoma subtypes with clear differences in prognosis. We characterize the subtype with a poorer disease-specific survival as Immune-infiltrated as it shows a higher abundance of immune cells within tumors. In comparison, the subtype with a better survival, termed the Cellular type, has a higher tumor cellularity and an enrichment of extracellular matrix and cell-to-cell interaction pathways that are hypomethylated at gene promoters. More notably, we demonstrate that plasma methylome-based non-invasive biomarkers can be used for chordoma diagnosis and prognostic subtyping, which can transform patient treatment by guiding surgical planning decisions. Accordingly, individualized treatment approaches depending on prognostic subtype may balance aggressiveness in extent of resection with the risk for treatment-induced neurological deficits. Introduction: The identification of molecular tumor subtypes has changed management approaches for many cancers. Bone and central nervous system (CNS) tumors remain an exception, to a large extent, including chordomas. Although chordomas are rare bone sarcomas of the skull-base and spine comprising 1-4% of primary aggressive bone cancers, they cause devastating quality of life impacts due to neurological morbidities, metastasize to other organs in 30-40%, and have a relatively high mortality with 10-year survival being 40%.1,2 Despite treatment with surgery and radiotherapy according to global consensus guidelines,1,3 outcomes range extensively with up to 10% surviving under 1 year and one-third living over 20 years.2 There are currently no major prognostic factors to identify high-risk patients histopathologically or clinically, apart from treatment details including extent of resection and quality of radiotherapy.1,3 Genomic4 and transcriptomic5 studies have not identified prognostic biomarkers. DNA methylation signatures can accurately diagnose CNS tumors6 and also prognosticate meningiomas of the CNS.7 There are a few existing chordoma methylation studies, however, with very small sample sizes and they do not resolve prognostic subtypes.6,8,9 Here, using samples from multi-institutional sources to generate a larger cohort of chordomas, we identified Translational Research 2 methylation-based prognostic subtypes that may allow clinicians to tailor treatment aggressiveness to patient risk. Furthermore, identifying non-invasive preoperative biomarkers for chordoma diagnosis and prognostication could transform treatment by guiding surgical planning to balance aggressiveness in extent of surgical resection with morbidities associated with loss of neurological function.1 Here, we also leverage our group’s novel liquid biopsy approach to explore whether plasma methylated circulating cell-free tumor DNA (cfDNA) obtained by immunoprecipitation and high-throughput sequencing (cfMeDIP-seq) can serve as a reliable biomarker.10-13 Methods: DNA methylation profiles were obtained using the Illumina EPIC array on 68 chordoma samples collected over a 22-year period with comprehensive clinical annotation. Consensus clustering of tumors using the top 15,000 most variably methylated CpG sites was performed in the full cohort and in a randomly separated training set (N=37). The most variably methylated CpGs in the training subcohort were used for consensus clustering of the independent testing subcohort for validation. A multivariate Cox analysis with all prognostic variables was performed. Gene-set enrichment analyses (GSEA) assessed genes differentially methylated at promoters to characterize methylation clusters and compare tumors by location. A cell deconvolution analysis using methylCIBERSORT assessed tumor immune cell compositions. Leukocytes unmethylation for purity (LUMP) estimates evaluated tumor purity. A total of 36 plasma cfDNA methylomes were obtained from matched chordoma patients (with tumor methylation profiles) together with meningioma and spine metastasis patients that have clinical entities commonly included in chordoma differential diagnoses. Samples were randomly split into fifty 80% training and 20% testing sets. The top 300 differentially methylated regions (DMRs) in each pairwise comparison between tumor types in training sets were combined for multidimensional scaling (MDS) plotting. Fifty negative-binomial generalized linear models for each tumor type were built using DMRs derived from training sets. Models were evaluated in corresponding independent testing sets with areas under receiver operating characteristic curves (AUROC) to assess discriminative capacity in distinguishing tumor types.10,11 For representative cases having this clinical differential diagnosis, class probabilities from all oneclass-versus-other models that included the tumor of interest in the testing set were calculated to assess model accuracy. To evaluate whether plasma methylation signals are representative of the methylation patterns in chordoma tumors, Pearson’s correlation coefficients were computed for each patient between normalized cfMeDIP-seq read counts at each region and averaged EPIC array beta values for CpGs in the region. Results Prognostic methylation subtypes This cohort represents a range of clinical presentations with 55% females included, all adult ages represented (18-80 years), and a balance between skull-base (64%) and spinal (36%) tumor locations. Patient treatment included a gross-total resection (GTR) in 39%, subtotal resection (STR) in 61%, and adjuvant radiotherapy in 69%. Consensus clustering of tumors identified two stable clusters shown in Figure 1a; cluster 1 with a statistically significant poorer disease-specific survival than cluster 2 (Figure 1b, median 6.0 vs. 17.3 years, log-rank p=0.0062). Consensus clustering of both randomly divided training and testing sets using features derived in the training set identified prognostic clusters in each (Supplementary Figure 1a-b, training: p=0.011, testing: p=0.0081). In a multivariate analysis combining prognostic clinical factors with methylation-based clusters (Supplementary Figure 1c), cluster 1 (hazard ratio (HR)=16.5, 95% confidence interval (CI)=2.8-96.1, p=0.0018), subtotal resection (HR=8.9, 95% CI=1.2-67.1, p=0.0336), and non-receipt of adjuvant Translational Research 3 radiotherapy (HR=9.5, 95% CI=1.7-52.8, p=0.0103) all independently predicted poorer survival with statistical significance. Subtype characterization A GSEA of genes differentially methylated at promoters (Supplementary Figure 1d) identified significant pathways with hypomethylated gene promoters, typically resulting in transcription, for each cluster (Figure 1c). Cluster 1 pathways were mainly immuneand transcription/translation-related while those in cluster 2 included cell-to-cell interaction, extracellular matrix, and angiogenesis pathways. Accordingly, cluster 1 was termed the Immune-infiltrated subtype and cluster 2 the Cellular subtype. Although tumor location was not predictive of disease-specific survival (univariate Cox p=0.1593), the GSEA in Supplementary Figure 1e revealed pathways with treatment-related implications14 including both kinase activity (encompassing PDGFR and KIT genes) and vascular proliferation hypomethylated pathways that are enriched in spinal chordomas. A greater abundance of neutrophils (7.0 fold, p<0.0001), B lymphocytes (2.5 fold, p=0.002), and natural killer cells (1.6 fold, p=0.045) was observed in cluster 1 chordomas (Figure 1d). Cytotoxic T lymphocytes with known anti-tumor activity15 were not differentially abundant between subtypes (Supplementary Figure 1f). Furthermore, tumor purity estimates in Figure 1e are higher in cluster 2, supporting the Cellular nature of this subtype (median 0.66 vs. 0.45, p<0.0001). Non-invasive diagnosis and subtyping An MDS plot of DMRs between chordomas, meningiomas, and spine metastases obtained from plasma cfDNA methylomes depicts class separation of chordomas from representative clinical differential diagnoses (Figure 2a). The fifty iterations of chordoma-versus-other models differentiated chordomas from meningiomas and spinal metastases with a high discriminative capacity in testing sets (mean AUROC=0.84, 95%CI=0.52-1.00) as shown in Figure 2b. A high correlation between chordoma tumor tissue methylation values and plasma cfMeDIP-seq signals was observed. Figure 2c displays a representative correlation plot and Pearson coefficients from all correlation plots portrayed in Figure 2d show high tumor-to-plasma correlations for both Immune-infiltrated (median r=0.69, 95%CI=0.66-0.72, p<2.2x10-16 for all) and Cellular chordomas (median r=0.67, 95%CI=0.62-0.72, p<2.2x10-16 for all). The top 7000 DMRs between subtypes distinguish them by hierarchical clustering (Figure 2e). We illustrate two clinical cases in Figure 2f-g where two expert neuroradiologists provided top differential diagnoses of a skull-base meningioma (f) and a spinal metastasis (g). Histopathology confirmed chordoma diagnoses in these patients and the cfMeDIP-seq based models accurately diagnosed both cases. Discussion: In this study we demonstrate, for the first time, two distinct methylation subtypes of chordoma, which we term Immune-infiltrated and Cellular. These subtypes have prognostic value independent of clinical factors and resolve the range of chordoma patient outcomes that are not explained by clinical, genomic, or transcriptomic features. We show that non-invasive diagnosis and prognostic subtyping of chordomas using plasma methylomes is possible with high accuracy and a discriminative capacity comparable to what we have shown previously for non-invasive identification of other cancers.10,11 We believe non-invasive chordoma prognostication will t