Pan-cancer proteogenomics expands the landscape of therapeutic targets.

Fewer than 200 proteins are targeted by cancer drugs approved by the Food and Drug Administration (FDA). We integrate Clinical Proteomic Tumor Analysis Consortium (CPTAC) proteogenomics data from 1,043 patients across 10 cancer types with additional public datasets to identify potential therapeutic targets. Pan-cancer analysis of 2,863 druggable proteins reveals a wide abundance range and identifies biological factors that affect mRNA-protein correlation. Integration of proteomic data from tumors and genetic screen data from cell lines identifies protein overexpression- or hyperactivation-driven druggable dependencies, enabling accurate predictions of effective drug targets. Proteogenomic identification of synthetic lethality provides a strategy to target tumor suppressor gene loss. Combining proteogenomic analysis and MHC binding prediction prioritizes mutant KRAS peptides as promising public neoantigens. Computational identification of shared tumor-associated antigens followed by experimental confirmation nominates peptides as immunotherapy targets. These analyses, summarized at https://targets.linkedomics.org, form a comprehensive landscape of protein and peptide targets for companion diagnostics, drug repurposing, and therapy development.

Proteogenomic analysis of chemo-refractory high-grade serous ovarian cancer.

To improve the understanding of chemo-refractory high-grade serous ovarian cancers (HGSOCs), we characterized the proteogenomic landscape of 242 (refractory and sensitive) HGSOCs, representing one discovery and two validation cohorts across two biospecimen types (formalin-fixed paraffin-embedded and frozen). We identified a 64-protein signature that predicts with high specificity a subset of HGSOCs refractory to initial platinum-based therapy and is validated in two independent patient cohorts. We detected significant association between lack of Ch17 loss of heterozygosity (LOH) and chemo-refractoriness. Based on pathway protein expression, we identified 5 clusters of HGSOC, which validated across two independent patient cohorts and patient-derived xenograft (PDX) models. These clusters may represent different mechanisms of refractoriness and implicate putative therapeutic vulnerabilities.

A proteogenomic portrait of lung squamous cell carcinoma.

Lung squamous cell carcinoma (LSCC) remains a leading cause of cancer death with few therapeutic options. We characterized the proteogenomic landscape of LSCC, providing a deeper exposition of LSCC biology with potential therapeutic implications. We identify NSD3 as an alternative driver in FGFR1-amplified tumors and low-p63 tumors overexpressing the therapeutic target survivin. SOX2 is considered undruggable, but our analyses provide rationale for exploring chromatin modifiers such as LSD1 and EZH2 to target SOX2-overexpressing tumors. Our data support complex regulation of metabolic pathways by crosstalk between post-translational modifications including ubiquitylation. Numerous immune-related proteogenomic observations suggest directions for further investigation. Proteogenomic dissection of CDKN2A mutations argue for more nuanced assessment of RB1 protein expression and phosphorylation before declaring CDK4/6 inhibition unsuccessful. Finally, triangulation between LSCC, LUAD, and HNSCC identified both unique and common therapeutic vulnerabilities. These observations and proteogenomics data resources may guide research into the biology and treatment of LSCC.

Proteogenomic characterization of pancreatic ductal adenocarcinoma.

Pancreatic ductal adenocarcinoma (PDAC) is a highly aggressive cancer with poor patient survival. Toward understanding the underlying molecular alterations that drive PDAC oncogenesis, we conducted comprehensive proteogenomic analysis of 140 pancreatic cancers, 67 normal adjacent tissues, and 9 normal pancreatic ductal tissues. Proteomic, phosphoproteomic, and glycoproteomic analyses were used to characterize proteins and their modifications. In addition, whole-genome sequencing, whole-exome sequencing, methylation, RNA sequencing (RNA-seq), and microRNA sequencing (miRNA-seq) were performed on the same tissues to facilitate an integrated proteogenomic analysis and determine the impact of genomic alterations on protein expression, signaling pathways, and post-translational modifications. To ensure robust downstream analyses, tumor neoplastic cellularity was assessed via multiple orthogonal strategies using molecular features and verified via pathological estimation of tumor cellularity based on histological review. This integrated proteogenomic characterization of PDAC will serve as a valuable resource for the community, paving the way for early detection and identification of novel therapeutic targets.

Proteogenomic Characterization Reveals Therapeutic Vulnerabilities in Lung Adenocarcinoma.

To explore the biology of lung adenocarcinoma (LUAD) and identify new therapeutic opportunities, we performed comprehensive proteogenomic characterization of 110 tumors and 101 matched normal adjacent tissues (NATs) incorporating genomics, epigenomics, deep-scale proteomics, phosphoproteomics, and acetylproteomics. Multi-omics clustering revealed four subgroups defined by key driver mutations, country, and gender. Proteomic and phosphoproteomic data illuminated biology downstream of copy number aberrations, somatic mutations, and fusions and identified therapeutic vulnerabilities associated with driver events involving KRAS, EGFR, and ALK. Immune subtyping revealed a complex landscape, reinforced the association of STK11 with immune-cold behavior, and underscored a potential immunosuppressive role of neutrophil degranulation. Smoking-associated LUADs showed correlation with other environmental exposure signatures and a field effect in NATs. Matched NATs allowed identification of differentially expressed proteins with potential diagnostic and therapeutic utility. This proteogenomics dataset represents a unique public resource for researchers and clinicians seeking to better understand and treat lung adenocarcinomas.

Proteogenomic Characterization of Endometrial Carcinoma.

We undertook a comprehensive proteogenomic characterization of 95 prospectively collected endometrial carcinomas, comprising 83 endometrioid and 12 serous tumors. This analysis revealed possible new consequences of perturbations to the p53 and Wnt/ß-catenin pathways, identified a potential role for circRNAs in the epithelial-mesenchymal transition, and provided new information about proteomic markers of clinical and genomic tumor subgroups, including relationships to known druggable pathways. An extensive genome-wide acetylation survey yielded insights into regulatory mechanisms linking Wnt signaling and histone acetylation. We also characterized aspects of the tumor immune landscape, including immunogenic alterations, neoantigens, common cancer/testis antigens, and the immune microenvironment, all of which can inform immunotherapy decisions. Collectively, our multi-omic analyses provide a valuable resource for researchers and clinicians, identify new molecular associations of potential mechanistic significance in the development of endometrial cancers, and suggest novel approaches for identifying potential therapeutic targets.

Proteogenomic Analysis of Human Colon Cancer Reveals New Therapeutic Opportunities.

We performed the first proteogenomic study on a prospectively collected colon cancer cohort. Comparative proteomic and phosphoproteomic analysis of paired tumor and normal adjacent tissues produced a catalog of colon cancer-associated proteins and phosphosites, including known and putative new biomarkers, drug targets, and cancer/testis antigens. Proteogenomic integration not only prioritized genomically inferred targets, such as copy-number drivers and mutation-derived neoantigens, but also yielded novel findings. Phosphoproteomics data associated Rb phosphorylation with increased proliferation and decreased apoptosis in colon cancer, which explains why this classical tumor suppressor is amplified in colon tumors and suggests a rationale for targeting Rb phosphorylation in colon cancer. Proteomics identified an association between decreased CD8 T cell infiltration and increased glycolysis in microsatellite instability-high (MSI-H) tumors, suggesting glycolysis as a potential target to overcome the resistance of MSI-H tumors to immune checkpoint blockade. Proteogenomics presents new avenues for biological discoveries and therapeutic development.

IDPpub: Illuminating the Dark Phosphoproteome Through PubMed Mining.

Global phosphoproteomics experiments quantify tens of thousands of phosphorylation sites. However, data interpretation is hampered by our limited knowledge on functions, biological contexts, or precipitating enzymes of the phosphosites. This study establishes a repository of phosphosites with associated evidence in biomedical abstracts, using deep learning-based natural language processing techniques. Our model for illuminating the dark phosphoproteome through PubMed mining (IDPpub) was generated by fine-tuning BioBERT, a deep learning tool for biomedical text mining. Trained using sentences containing protein substrates and phosphorylation site positions from 3000 abstracts, the IDPpub model was then used to extract phosphorylation sites from all MEDLINE abstracts. The extracted proteins were normalized to gene symbols using the National Center for Biotechnology Information gene query, and sites were mapped to human UniProt sequences using ProtMapper and mouse UniProt sequences by direct match. Precision and recall were calculated using 150 curated abstracts, and utility was assessed by analyzing the CPTAC (Clinical Proteomics Tumor Analysis Consortium) pan-cancer phosphoproteomics datasets and the PhosphoSitePlus database. Using 10-fold cross validation, pairs of correct substrates and phosphosite positions were extracted with an average precision of 0.93 and recall of 0.94. After entity normalization and site mapping to human reference sequences, an independent validation achieved a precision of 0.91 and recall of 0.77. The IDPpub repository contains 18,458 unique human phosphorylation sites with evidence sentences from 58,227 abstracts and 5918 mouse sites in 14,610 abstracts. This included evidence sentences for 1803 sites identified in CPTAC studies that are not covered by manually curated functional information in PhosphoSitePlus. Evaluation results demonstrate the potential of IDPpub as an effective biomedical text mining tool for collecting phosphosites. Moreover, the repository (http://idppub.ptmax.org), which can be automatically updated, can serve as a powerful complement to existing resources.

Frozen tissue coring and layered histological analysis improves cell type-specific proteogenomic characterization of pancreatic adenocarcinoma.

BACKGROUND: Omics characterization of pancreatic adenocarcinoma tissue is complicated by the highly heterogeneous and mixed populations of cells. We evaluate the feasibility and potential benefit of using a coring method to enrich specific regions from bulk tissue and then perform proteogenomic analyses. METHODS: We used the Biopsy Trifecta Extraction (BioTExt) technique to isolate cores of epithelial-enriched and stroma-enriched tissue from pancreatic tumor and adjacent tissue blocks. Histology was assessed at multiple depths throughout each core. DNA sequencing, RNA sequencing, and proteomics were performed on the cored and bulk tissue samples. Supervised and unsupervised analyses were performed based on integrated molecular and histology data. RESULTS: Tissue cores had mixed cell composition at varying depths throughout. Average cell type percentages assessed by histology throughout the core were better associated with KRAS variant allele frequencies than standard histology assessment of the cut surface. Clustering based on serial histology data separated the cores into three groups with enrichment of neoplastic epithelium, stroma, and acinar cells, respectively. Using this classification, tumor overexpressed proteins identified in bulk tissue analysis were assigned into epithelial- or stroma-specific categories, which revealed novel epithelial-specific tumor overexpressed proteins. CONCLUSIONS: Our study demonstrates the feasibility of multi-omics data generation from tissue cores, the necessity of interval H&E stains in serial histology sections, and the utility of coring to improve analysis over bulk tissue data.

Hemin-Induced Death Models Hemorrhagic Stroke and Is a Variant of Classical Neuronal Ferroptosis.

Ferroptosis is a caspase-independent, iron-dependent form of regulated necrosis extant in traumatic brain injury, Huntington disease, and hemorrhagic stroke. It can be activated by cystine deprivation leading to glutathione depletion, the insufficiency of the antioxidant glutathione peroxidase-4, and the hemolysis products hemoglobin and hemin. A cardinal feature of ferroptosis is extracellular signal-regulated kinase (ERK)1/2 activation culminating in its translocation to the nucleus. We have previously confirmed that the mitogen-activated protein (MAP) kinase kinase (MEK) inhibitor U0126 inhibits persistent ERK1/2 phosphorylation and ferroptosis. Here, we show that hemin exposure, a model of secondary injury in brain hemorrhage and ferroptosis, activated ERK1/2 in mouse neurons. Accordingly, MEK inhibitor U0126 protected against hemin-induced ferroptosis. Unexpectedly, U0126 prevented hemin-induced ferroptosis independent of its ability to inhibit ERK1/2 signaling. In contrast to classical ferroptosis in neurons or cancer cells, chemically diverse inhibitors of MEK did not block hemin-induced ferroptosis, nor did the forced expression of the ERK-selective MAP kinase phosphatase (MKP)3. We conclude that hemin or hemoglobin-induced ferroptosis, unlike glutathione depletion, is ERK1/2-independent. Together with recent studies, our findings suggest the existence of a novel subtype of neuronal ferroptosis relevant to bleeding in the brain that is 5-lipoxygenase-dependent, ERK-independent, and transcription-independent. Remarkably, our unbiased phosphoproteome analysis revealed dramatic differences in phosphorylation induced by two ferroptosis subtypes. As U0126 also reduced cell death and improved functional recovery after hemorrhagic stroke in male mice, our analysis also provides a template on which to build a search for U0126's effects in a variant of neuronal ferroptosis.SIGNIFICANCE STATEMENT Ferroptosis is an iron-dependent mechanism of regulated necrosis that has been linked to hemorrhagic stroke. Common features of ferroptotic death induced by diverse stimuli are the depletion of the antioxidant glutathione, production of lipoxygenase-dependent reactive lipids, sensitivity to iron chelation, and persistent activation of extracellular signal-regulated kinase (ERK) signaling. Unlike classical ferroptosis induced in neurons or cancer cells, here we show that ferroptosis induced by hemin is ERK-independent. Paradoxically, the canonical MAP kinase kinase (MEK) inhibitor U0126 blocks brain hemorrhage-induced death. Altogether, these data suggest that a variant of ferroptosis is unleashed in hemorrhagic stroke. We present the first, unbiased phosphoproteomic analysis of ferroptosis as a template on which to understand distinct paths to cell death that meet the definition of ferroptosis.

Graph Algorithms for Condensing and Consolidating Gene Set Analysis Results.

Gene set analysis plays a critical role in the functional interpretation of omics data. Although this is typically done for one omics experiment at a time, there is an increasing need to combine gene set analysis results from multiple experiments performed on the same or different omics platforms, such as in multi-omics studies. Integrating results from multiple experiments is challenging, and annotation redundancy between gene sets further obscures clear conclusions. We propose to use a weighted set cover algorithm to reduce redundancy of gene sets identified in a single experiment. Next, we use affinity propagation to consolidate similar gene sets identified from multiple experiments into clusters and to automatically determine the most representative gene set for each cluster. Using three examples from over representation analysis and gene set enrichment analysis, we showed that weighted set cover outperformed a previously published set cover method and reduced the number of gene sets by 52-77%. Focusing on overlapping genes between the list of input genes and the enriched gene sets in over-representation analysis and leading-edge genes in gene set enrichment analysis further reduced the number of gene sets. A use case combining enrichment analysis results from RNA-Seq and proteomics data comparing basal and luminal A breast cancer samples highlighted the known difference in proliferation and DNA damage response. Finally, we used these algorithms for a pan-cancer survival analysis. Our analysis clearly revealed prognosis-related pathways common to multiple cancer types or specific to individual cancer types, as well as pathways associated with prognosis in different directions in different cancer types. We implemented these two algorithms in an R package, Sumer, which generates tables and static and interactive plots for exploration and publication. Sumer is publicly available at https://github.com/bzhanglab/sumer.

Deep Learning in Proteomics.

Proteomics, the study of all the proteins in biological systems, is becoming a data-rich science. Protein sequences and structures are comprehensively catalogued in online databases. With recent advancements in tandem mass spectrometry (MS) technology, protein expression and post-translational modifications (PTMs) can be studied in a variety of biological systems at the global scale. Sophisticated computational algorithms are needed to translate the vast amount of data into novel biological insights. Deep learning automatically extracts data representations at high levels of abstraction from data, and it thrives in data-rich scientific research domains. Here, a comprehensive overview of deep learning applications in proteomics, including retention time prediction, MS/MS spectrum prediction, de novo peptide sequencing, PTM prediction, major histocompatibility complex-peptide binding prediction, and protein structure prediction, is provided. Limitations and the future directions of deep learning in proteomics are also discussed. This review will provide readers an overview of deep learning and how it can be used to analyze proteomics data.

Using phosphoproteomics data to understand cellular signaling: a comprehensive guide to bioinformatics resources.

Mass spectrometry-based phosphoproteomics is becoming an essential methodology for the study of global cellular signaling. Numerous bioinformatics resources are available to facilitate the translation of phosphopeptide identification and quantification results into novel biological and clinical insights, a critical step in phosphoproteomics data analysis. These resources include knowledge bases of kinases and phosphatases, phosphorylation sites, kinase inhibitors, and sequence variants affecting kinase function, and bioinformatics tools that can predict phosphorylation sites in addition to the kinase that phosphorylates them, infer kinase activity, and predict the effect of mutations on kinase signaling. However, these resources exist in silos and it is challenging to select among multiple resources with similar functions. Therefore, we put together a comprehensive collection of resources related to phosphoproteomics data interpretation, compared the use of tools with similar functions, and assessed the usability from the standpoint of typical biologists or clinicians. Overall, tools could be improved by standardization of enzyme names, flexibility of data input and output format, consistent maintenance, and detailed manuals.

The peroxisome proliferator-activated receptor-ß/δ antagonist GSK0660 mitigates retinal cell inflammation and leukostasis.

Diabetic retinopathy (DR) is triggered by retinal cell damage stimulated by the diabetic milieu, including increased levels of intraocular free fatty acids. Free fatty acids may serve as an initiator of inflammatory cytokine release from Müller cells, and the resulting cytokines are potent stimulators of retinal endothelial pathology, such as leukostasis, vascular permeability, and basement membrane thickening. Our previous studies have elucidated a role for peroxisome proliferator-activated receptor-ß/δ (PPARß/δ) in promoting several steps in the pathologic cascade in DR, including angiogenesis and expression of inflammatory mediators. Furthermore, PPARß/δ is a known target of lipid signaling, suggesting a potential role for this transcription factor in fatty acid-induced retinal inflammation. Therefore, we hypothesized that PPARß/δ stimulates both the induction of inflammatory mediators by Müller cells as well the paracrine induction of leukostasis in endothelial cells (EC) by Müller cell inflammatory products. To test this, we used the PPARß/δ inhibitor, GSK0660, in primary human Müller cells (HMC), human retinal microvascular endothelial cells (HRMEC) and mouse retina. We found that palmitic acid (PA) activation of PPARß/δ in HMC leads to the production of pro-angiogenic and/or inflammatory cytokines that may constitute DR-relevant upstream paracrine inflammatory signals to EC and other retinal cells. Downstream, EC transduce these signals and increase their synthesis and release of chemokines such as CCL8 and CXCL10 that regulate leukostasis and other cellular events related to vascular inflammation in DR. Our results indicate that PPARß/δ inhibition mitigates these upstream (MC) as well as downstream (EC) inflammatory signaling events elicited by metabolic stimuli and inflammatory cytokines. Therefore, our data suggest that PPARß/δ inhibition is a potential therapeutic strategy against early DR pathology.

RNA-seq identifies a role for the PPARß/δ inverse agonist GSK0660 in the regulation of TNFα-induced cytokine signaling in retinal endothelial cells.

PURPOSE: The peroxisome proliferator-activated receptor beta/delta (PPARß/δ) is a transcription factor with roles in metabolism, angiogenesis, and inflammation. It has yet undefined roles in retinal inflammation and diabetic retinopathy (DR). We used RNA-seq to better understand the role of the antagonist and inverse agonist of PPARß/δ, GSK0660, in TNFα-induced inflammation. Understanding the underlying mechanisms of vascular inflammation could lead to new treatments for DR. METHODS: RNA was isolated from human retinal microvascular endothelial cells treated with a vehicle, TNFα, or TNFα plus GSK0660. RNA-seq was performed with a 50 bp single read protocol. The differential expression was determined using edgeR and gene ontology, and a pathway analysis was performed using DAVID. RNA-seq validation was performed using qRT-PCR using the primers for ANGPTL4, CCL8, NOV, CXCL10, and PDPK1. RESULTS: TNFα differentially regulated 1,830 transcripts, many of which are involved in the cytokine-cytokine receptor interaction, chemokine signaling, and inflammatory response. Additionally, TNFα highly upregulated genes involved in leukocyte recruitment, including CCL5, CX3CL1, and CXCL10. GSK0660 differentially regulated 273 transcripts in TNFα-treated cells compared to TNFα alone. A pathway analysis revealed the enrichment of cytokine-cytokine receptor signaling. In particular, GSK0660 blocks the TNFα-induced upregulation of CCL8, a chemokine involved in leukocyte recruitment. CONCLUSIONS: TNFα regulates several genes related to retinal leukostasis in retinal endothelial cells. GSK0660 blocks the effect of TNFα on the expressions of cytokines involved in leukocyte recruitment, including CCL8, CCL17, and CXCL10 and it may therefore block TNFα-induced retinal leukostasis.

Hypoxia-induced expression of VEGF splice variants and protein in four retinal cell types.

The purpose of this study was to investigate the hypoxia-induced Vegf120, Vegf164 and Vegf188 mRNA expression profiles in rat Müller cells (MC), astrocytes, retinal pigmented epithelial cells (RPE) and retinal microvascular endothelial cells (RMEC) and correlate these findings to VEGF secreted protein. Cultured cells were exposed to normoxia or hypoxia. Total RNA was isolated from cell lysates and Vegf splice variant mRNA copy numbers were assayed by a validated qRT-PCR external calibration curve method. mRNA copy numbers were normalized to input total RNA. Conditioned medium was collected from cells and assayed for total VEGF protein by ELISA. Hypoxia increased total Vegf mRNA and secreted protein in all the retinal cell types, with the highest levels observed in MC and astrocytes ranking second. Total Vegf mRNA levels in hypoxic RPE and RMEC were comparable; however, the greatest hypoxic induction of each Vegf splice variant mRNA was observed in RMEC. RPE and RMEC ranked 3rd and 4th respectively, in terms of secreted total VEGF protein in hypoxia. The Vegf120, Vegf164 and Vegf188 mRNA splice variants were all increased in hypoxic cells compared to normoxic controls. In normoxia, the relative Vegf splice variant mRNA levels ranked from highest to lowest for each cell type were Vegf164 > Vegf120 > Vegf188. Hypoxic induction did not alter this ranking, although it did favor an increased stoichiometry of Vegf164 mRNA over the other two splice variants. MC and astrocytes are likely to be the major sources of total Vegf, Vegf164 splice variant mRNAs, and VEGF protein in retinal hypoxia.

A proteogenomics data-driven knowledge base of human cancer.

By combining mass-spectrometry-based proteomics and phosphoproteomics with genomics, epi-genomics, and transcriptomics, proteogenomics provides comprehensive molecular characterization of cancer. Using this approach, the Clinical Proteomic Tumor Analysis Consortium (CPTAC) has characterized over 1,000 primary tumors spanning 10 cancer types, many with matched normal tissues. Here, we present LinkedOmicsKB, a proteogenomics data-driven knowledge base that makes consistently processed and systematically precomputed CPTAC pan-cancer proteogenomics data available to the public through ∼40,000 gene-, protein-, mutation-, and phenotype-centric web pages. Visualization techniques facilitate efficient exploration and reasoning of complex, interconnected data. Using three case studies, we illustrate the practical utility of LinkedOmicsKB in providing new insights into genes, phosphorylation sites, somatic mutations, and cancer phenotypes. With precomputed results of 19,701 coding genes, 125,969 phosphosites, and 256 genotypes and phenotypes, LinkedOmicsKB provides a comprehensive resource to accelerate proteogenomics data-driven discoveries to improve our understanding and treatment of human cancer. A record of this paper's transparent peer review process is included in the supplemental information.