Stromal Signals Dominate Gene Expression Signature Scores That Aim to Describe Cancer Cell-intrinsic Stemness or Mesenchymality Characteristics.

Epithelial-to-mesenchymal transition (EMT) in cancer cells confers migratory abilities, a crucial aspect in the metastasis of tumors that frequently leads to death. In multiple studies, authors proposed gene expression signatures for EMT, stemness, or mesenchymality of tumors based on bulk tumor expression profiling. However, recent studies suggested that noncancerous cells from the microenvironment or macroenvironment heavily influence such signature profiles. Here, we strengthen these findings by investigating 11 published and frequently referenced gene expression signatures that were proposed to describe EMT-related (EMT, mesenchymal, or stemness) characteristics in various cancer types. By analyses of bulk, single-cell, and pseudobulk expression data, we show that the cell type composition of a tumor sample frequently dominates scores of these EMT-related signatures. A comprehensive, integrated analysis of bulk RNA sequencing (RNA-seq) and single-cell RNA-seq data shows that stromal cells, most often fibroblasts, are the main drivers of EMT-related signature scores. We call attention to the risk of false conclusions about tumor properties when interpreting EMT-related signatures, especially in a clinical setting: high patient scores of EMT-related signatures or calls of "stemness subtypes" often result from low cancer cell content in tumor biopsies rather than cancer cell-specific stemness or mesenchymal/EMT characteristics. SIGNIFICANCE: Cancer self-renewal and migratory abilities are often characterized via gene module expression profiles, also called EMT or stemness gene expression signatures. Using published clinical tumor samples, cancer cell lines, and single cancer cells, we highlight the dominating influence of noncancer cells in low cancer cell content biopsies on their scores. We caution on their application for low cancer cell content clinical cancer samples with the intent to assign such characteristics or subtypes.

Immune cell type signature discovery and random forest classification for analysis of single cell gene expression datasets.

Background: Robust immune cell gene expression signatures are central to the analysis of single cell studies. Nearly all known sets of immune cell signatures have been derived by making use of only single gene expression datasets. Utilizing the power of multiple integrated datasets could lead to high-quality immune cell signatures which could be used as superior inputs to machine learning-based cell type classification approaches. Results: We established a novel workflow for the discovery of immune cell type signatures based primarily on gene-versus-gene expression similarity. It leverages multiple datasets, here seven single cell expression datasets from six different cancer types and resulted in eleven immune cell type-specific gene expression signatures. We used these to train random forest classifiers for immune cell type assignment for single-cell RNA-seq datasets. We obtained similar or better prediction results compared to commonly used methods for cell type assignment in independent benchmarking datasets. Our gene signature set yields higher prediction scores than other published immune cell type gene sets in random forest-based cell type classification. We further demonstrate how our approach helps to avoid bias in downstream statistical analyses by re-analysis of a published IFN stimulation experiment. Discussion and conclusion: We demonstrated the quality of our immune cell signatures and their strong performance in a random forest-based cell typing approach. We argue that classifying cells based on our comparably slim sets of genes accompanied by a random forest-based approach not only matches or outperforms widely used published approaches. It also facilitates unbiased downstream statistical analyses of differential gene expression between cell types for significantly more genes compared to previous cell classification algorithms.

Synthetic lethality prediction in DNA damage repair, chromatin remodeling and the cell cycle using multi-omics data from cell lines and patients.

Discovering synthetic lethal (SL) gene partners of cancer genes is an important step in developing cancer therapies. However, identification of SL interactions is challenging, due to a large number of possible gene pairs, inherent noise and confounding factors in the observed signal. To discover robust SL interactions, we devised SLIDE-VIP, a novel framework combining eight statistical tests, including a new patient data-based test iSurvLRT. SLIDE-VIP leverages multi-omics data from four different sources: gene inactivation cell line screens, cancer patient data, drug screens and gene pathways. We applied SLIDE-VIP to discover SL interactions between genes involved in DNA damage repair, chromatin remodeling and cell cycle, and their potentially druggable partners. The top 883 ranking SL candidates had strong evidence in cell line and patient data, 250-fold reducing the initial space of 200K pairs. Drug screen and pathway tests provided additional corroboration and insights into these interactions. We rediscovered well-known SL pairs such as RB1 and E2F3 or PRKDC and ATM, and in addition, proposed strong novel SL candidates such as PTEN and PIK3CB. In summary, SLIDE-VIP opens the door to the discovery of SL interactions with clinical potential. All analysis and visualizations are available via the online SLIDE-VIP WebApp.

CONET: copy number event tree model of evolutionary tumor history for single-cell data.

Copy number alterations constitute important phenomena in tumor evolution. Whole genome single-cell sequencing gives insight into copy number profiles of individual cells, but is highly noisy. Here, we propose CONET, a probabilistic model for joint inference of the evolutionary tree on copy number events and copy number calling. CONET employs an efficient, regularized MCMC procedure to search the space of possible model structures and parameters. We introduce a range of model priors and penalties for efficient regularization. CONET reveals copy number evolution in two breast cancer samples, and outperforms other methods in tree reconstruction, breakpoint identification and copy number calling.

RosettaSX: Reliable gene expression signature scoring of cancer models and patients.

Gene expression signatures have proven their potential to characterize important cancer phenomena like oncogenic signaling pathway activities, cellular origins of tumors, or immune cell infiltration into tumor tissues. Large collections of expression signatures provide the basis for their application to data sets, but the applicability of each signature in a new experimental context must be reassessed. We apply a methodology that utilizes the previously developed concept of coherent expression of genes in signatures to identify translatable signatures before scoring their activity in single tumors. We present a web interface (www.rosettasx.com) that applies our methodology to expression data from the Cancer Cell Line Encyclopaedia and The Cancer Genome Atlas. Configurable heat maps visualize per-cancer signature scores for 293 hand-curated literature-derived gene sets representing a wide range of cancer-relevant transcriptional modules and phenomena. The platform allows users to complement heatmaps of signature scores with molecular information on SNVs, CNVs, gene expression, gene dependency, and protein abundance or to analyze own signatures. Clustered heatmaps and further plots to drill-down results support users in studying oncological processes in cancer subtypes, thereby providing a rich resource to explore how mechanisms of cancer interact with each other as demonstrated by exemplary analyses of 2 cancer types.

Interpretable deep recommender system model for prediction of kinase inhibitor efficacy across cancer cell lines.

Computational models for drug sensitivity prediction have the potential to significantly improve personalized cancer medicine. Drug sensitivity assays, combined with profiling of cancer cell lines and drugs become increasingly available for training such models. Multiple methods were proposed for predicting drug sensitivity from cancer cell line features, some in a multi-task fashion. So far, no such model leveraged drug inhibition profiles. Importantly, multi-task models require a tailored approach to model interpretability. In this work, we develop DEERS, a neural network recommender system for kinase inhibitor sensitivity prediction. The model utilizes molecular features of the cancer cell lines and kinase inhibition profiles of the drugs. DEERS incorporates two autoencoders to project cell line and drug features into 10-dimensional hidden representations and a feed-forward neural network to combine them into response prediction. We propose a novel interpretability approach, which in addition to the set of modeled features considers also the genes and processes outside of this set. Our approach outperforms simpler matrix factorization models, achieving R [Formula: see text] 0.82 correlation between true and predicted response for the unseen cell lines. The interpretability analysis identifies 67 biological processes that drive the cell line sensitivity to particular compounds. Detailed case studies are shown for PHA-793887, XMD14-99 and Dabrafenib.

Feature selection strategies for drug sensitivity prediction.

Drug sensitivity prediction constitutes one of the main challenges in personalized medicine. Critically, the sensitivity of cancer cells to treatment depends on an unknown subset of a large number of biological features. Here, we compare standard, data-driven feature selection approaches to feature selection driven by prior knowledge of drug targets, target pathways, and gene expression signatures. We asses these methodologies on Genomics of Drug Sensitivity in Cancer (GDSC) dataset, evaluating 2484 unique models. For 23 drugs, better predictive performance is achieved when the features are selected according to prior knowledge of drug targets and pathways. The best correlation of observed and predicted response using the test set is achieved for Linifanib (r = 0.75). Extending the drug-dependent features with gene expression signatures yields the most predictive models for 60 drugs, with the best performing example of Dabrafenib. For many compounds, even a very small subset of drug-related features is highly predictive of drug sensitivity. Small feature sets selected using prior knowledge are more predictive for drugs targeting specific genes and pathways, while models with wider feature sets perform better for drugs affecting general cellular mechanisms. Appropriate feature selection strategies facilitate the development of interpretable models that are indicative for therapy design.

Genetic Interactions and Tissue Specificity Modulate the Association of Mutations with Drug Response.

In oncology, biomarkers are widely used to predict subgroups of patients that respond to a given drug. Although clinical decisions often rely on single gene biomarkers, machine learning approaches tend to generate complex multi-gene biomarkers that are hard to interpret. Models predicting drug response based on multiple altered genes often assume that the effects of single alterations are independent. We asked whether the association of cancer driver mutations with drug response is modulated by other driver mutations or the tissue of origin. We developed an analytic framework based on linear regression to study interactions in pharmacogenomic data from two large cancer cell line panels. Starting from a model with only covariates, we included additional variables only if they significantly improved simpler models. This allows to systematically assess interactions in small, easily interpretable models. Our results show that including mutation-mutation interactions in drug response prediction models tends to improve model performance and robustness. For example, we found that TP53 mutations decrease sensitivity to BRAF inhibitors in BRAF-mutated cell lines and patient tumors, suggesting a therapeutic benefit of combining inhibition of oncogenic BRAF with reactivation of the tumor suppressor TP53. Moreover, we identified tissue-specific mutation-drug associations and synthetic lethal triplets where the simultaneous mutation of two genes sensitizes cells to a drug. In summary, our interaction-based approach contributes to a holistic view on the determining factors of drug response.

Association of EGFR expression level and cetuximab activity in patient-derived xenograft models of human non-small cell lung cancer.

PURPOSE: To explore in a panel of patient-derived xenograft models of human non-small cell lung cancer (NSCLC) whether high EGFR expression, was associated with cetuximab activity. EXPERIMENTAL DESIGN: NSCLC patient-derived xenograft models (n=45) were implanted subcutaneously into panels of nude mice and randomization cohorts were treated with either cetuximab, cisplatin, cisplatin plus cetuximab, vehicle control, or else were left untreated. Responses according to treatment were assessed at week 3 by analyzing the relative change in tumor volume and an experimental analogue of the Response Evaluation Criteria in Solid Tumors (RECIST) guidelines. An EGFR IHC score was calculated for each patient-derived xenograft model and response was assessed according to EGFR expression level. RESULTS: When tumors were stratified into high and low EGFR expression groups (IHC score threshold 200; scale 0-300), a stronger antitumor activity was seen in the high EGFR expression group compared with the low EGFR expression group in both the cetuximab monotherapy and cisplatin plus cetuximab combination therapy settings. For tumors treated with cisplatin plus cetuximab, the objective response rate was significantly higher in the high EGFR expression group compared with the low EGFR expression group (68% vs. 29%). Objective response rates were similar in high and low expression groups for tumors treated with cisplatin alone (27% vs. 24%, respectively). CONCLUSION: Cetuximab activity in NSCLC patient-derived xenograft models was demonstrated clearly only in tumors that expressed high levels of EGFR, as defined by an IHC score of ≥200.

An interferon response gene expression signature is activated in a subset of medulloblastomas.

Recent evidence suggests that cytomegalovirus infection contributes to the development of medulloblastomas. Differential activation of antiviral expression programs in medulloblastomas has not been investigated yet. In this study, we assess the relevance of an antiviral transcriptional response in medulloblastomas. We analyzed a gene expression signature of type I interferon response in three public gene expression data sets of medulloblastomas. Interferon response genes were found to be significantly coordinately regulated in two independent studies. We distilled a signature of 10 interferon response genes from two data sets. This signature exhibited strongly significant gene-versus-gene correlation of expression levels across samples in a third external medulloblastoma data set. Our medulloblastoma IFN signature identified a previously unrecognized patient subgroup partially overlapping the WNT and SHH subtypes proposed by others. We conclude that significant traces of differential activation of antiviral transcriptional response can be found in three independent medulloblastoma patient cohorts. This IFN activation signal often coincides with reduced proliferation scores. Our proposed 10-gene type I IFN response gene signature could help to assess antiviral states in further gene expression data sets of medulloblastomas or other cancers.

Female sex and estrogen receptor-beta attenuate cardiac remodeling and apoptosis in pressure overload.

We investigated sex differences and the role of estrogen receptor-beta (ERbeta) on myocardial hypertrophy in a mouse model of pressure overload. We performed transverse aortic constriction (TAC) or sham surgery in male and female wild-type (WT) and ERbeta knockout (ERbeta(-/-)) mice. All mice were characterized by echocardiography and hemodynamic measurements and were killed 9 wk after surgery. Left ventricular (LV) samples were analyzed by microarray profiling, real-time RT-PCR, and histology. After 9 wk, WT males showed more hypertrophy and heart failure signs than WT females. Notably, WT females developed a concentric form of hypertrophy, while males developed eccentric hypertrophy. ERbeta deletion augmented the TAC-induced increase in cardiomyocyte diameter in both sexes. Gene expression profiling revealed that WT male hearts had a stronger induction of matrix-related genes and a stronger repression of mitochondrial genes than WT female hearts. ERbeta(-/-) mice exhibited a different transcriptional response. ERbeta(-/-)/TAC mice of both sexes exhibited induction of proapoptotic genes with a stronger expression in ERbeta(-/-) males. Cardiac fibrosis was more pronounced in male WT/TAC than in female mice. This difference was abolished in ERbeta(-/-) mice. The number of apoptotic nuclei was increased in both sexes of ERbeta(-/-)/TAC mice, most prominent in males. Female sex offers protection against ventricular chamber dilation in the TAC model. Both female sex and ERbeta attenuate the development of fibrosis and apoptosis, thus slowing the progression to heart failure.

An expression module of WIPF1-coexpressed genes identifies patients with favorable prognosis in three tumor types.

Wiskott-Aldrich syndrome (WAS) predisposes patients to leukemia and lymphoma. WAS is caused by mutations in the protein WASP which impair its interaction with the WIPF1 protein. Here, we aim to identify a module of WIPF1-coexpressed genes and to assess its use as a prognostic signature for colorectal cancer, glioma, and breast cancer patients. Two public colorectal cancer microarray data sets were used for discovery and validation of the WIPF1 co-expression module. Based on expression of the WIPF1 signature, we classified more than 400 additional tumors with microarray data from our own experiments or from publicly available data sets according to their WIPF1 signature expression. This allowed us to separate patient populations for colorectal cancers, breast cancers, and gliomas for which clinical characteristics like survival times and times to relapse were analyzed. Groups of colorectal cancer, breast cancer, and glioma patients with low expression of the WIPF1 co-expression module generally had a favorable prognosis. In addition, the majority of WIPF1 signature genes are individually correlated with disease outcome in different studies. Literature gene network analysis revealed that among WIPF1 co-expressed genes known direct transcriptional targets of c-myc, ESR1 and p53 are enriched. The mean expression profile of WIPF1 signature genes is correlated with the profile of a proliferation signature. The WIPF1 signature is the first microarray-based prognostic expression signature primarily developed for colorectal cancer that is instrumental in other tumor types: low expression of the WIPF1 module is associated with better prognosis.

Genome-wide expression patterns of invasion front, inner tumor mass and surrounding normal epithelium of colorectal tumors.

Colorectal tumors have characteristic genome-wide expression patterns that allow their distinction from normal colon epithelia and facilitate clinical prognosis. The expression heterogeneity within a primary colorectal tumor has not been studied on a genome scale yet. Here we investigated three compartments of colorectal tumors, the invasion front, the inner tumor mass, and surrounding normal epithelial tissue by microdissection and microarray-based expression profiling. In both tumor compartments many genes were differentially expressed when compared to normal epithelium. The sets of significantly deregulated genes in both compartments overlapped to a large extent and revealed various interesting known and novel pathways that could have contributed to tumorigenesis. Cells from the invasion front and inner tumor mass, however, did not show significant differences in their expression profile, neither on the single gene level nor on the pathway level. Instead, gene expression differences between individuals are more pronounced as all patient-matched tumor samples clustered in close proximity to each other. With respect to invasion front and inner tumor mass we conclude that the specific tumor cell micro-environment does not have a strong influence on expression patterns: largely similar genome-wide expression programs operate in the invasion front and interior compartment of a colorectal tumor.

The highly conserved LepA is a ribosomal elongation factor that back-translocates the ribosome.

The ribosomal elongation cycle describes a series of reactions prolonging the nascent polypeptide chain by one amino acid and driven by two universal elongation factors termed EF-Tu and EF-G in bacteria. Here we demonstrate that the extremely conserved LepA protein, present in all bacteria and mitochondria, is a third elongation factor required for accurate and efficient protein synthesis. LepA has the unique function of back-translocating posttranslocational ribosomes, and the results suggest that it recognizes ribosomes after a defective translocation reaction and induces a back-translocation, thus giving EF-G a second chance to translocate the tRNAs correctly. We suggest renaming LepA as elongation factor 4 (EF4).

An inventory of yeast proteins associated with nucleolar and ribosomal components.

BACKGROUND: Although baker's yeast is a primary model organism for research on eukaryotic ribosome assembly and nucleoli, the list of its proteins that are functionally associated with nucleoli or ribosomes is still incomplete. We trained a naïve Bayesian classifier to predict novel proteins that are associated with yeast nucleoli or ribosomes based on parts lists of nucleoli in model organisms and large-scale protein interaction data sets. Phylogenetic profiling and gene expression analysis were carried out to shed light on evolutionary and regulatory aspects of nucleoli and ribosome assembly. RESULTS: We predict that, in addition to 439 known proteins, a further 62 yeast proteins are associated with components of the nucleolus or the ribosome. The complete set comprises a large core of archaeal-type proteins, several bacterial-type proteins, but mostly eukaryote-specific inventions. Expression of nucleolar and ribosomal genes tends to be strongly co-regulated compared to other yeast genes. CONCLUSION: The number of proteins associated with nucleolar or ribosomal components in yeast is at least 14% higher than known before. The nucleolus probably evolved from an archaeal-type ribosome maturation machinery by recruitment of several bacterial-type and mostly eukaryote-specific factors. Not only expression of ribosomal protein genes, but also expression of genes encoding the 90S processosome, are strongly co-regulated and both regulatory programs are distinct from each other.

A genome-wide map of aberrantly expressed chromosomal islands in colorectal cancer.

BACKGROUND: Cancer development is accompanied by genetic phenomena like deletion and amplification of chromosome parts or alterations of chromatin structure. It is expected that these mechanisms have a strong effect on regional gene expression. RESULTS: We investigated genome-wide gene expression in colorectal carcinoma (CRC) and normal epithelial tissues from 25 patients using oligonucleotide arrays. This allowed us to identify 81 distinct chromosomal islands with aberrant gene expression. Of these, 38 islands show a gain in expression and 43 a loss of expression. In total, 7.892 genes (25.3% of all human genes) are located in aberrantly expressed islands. Many chromosomal regions that are linked to hereditary colorectal cancer show deregulated expression. Also, many known tumor genes localize to chromosomal islands of misregulated expression in CRC. CONCLUSION: An extensive comparison with published CGH data suggests that chromosomal regions known for frequent deletions in colon cancer tend to show reduced expression. In contrast, regions that are often amplified in colorectal tumors exhibit heterogeneous expression patterns: even show a decrease of mRNA expression. Because for several islands of deregulated expression chromosomal aberrations have never been observed, we speculate that additional mechanisms (like abnormal states of regional chromatin) also have a substantial impact on the formation of co-expression islands in colorectal carcinoma.

Horizontal gene transfer in aminoacyl-tRNA synthetases including leucine-specific subtypes.

Aminoacyl-tRNA synthetases catalyze a fundamental reaction for the flow of genetic information from RNA to protein. Their presence in all organisms known today highlights their important role in the early evolution of life. We investigated the evolutionary history of aminoacyl-tRNA synthetases on the basis of sequence data from more than 200 Archaea, Bacteria, and Eukaryota. Phylogenetic profiles are in agreement with previous observations that many genes for aminoacyl-tRNA synthetases were transferred horizontally between species from all domains of life. We extended these findings by a detailed analysis of the history of leucyl-tRNA synthetases. Thereby, we identified a previously undetected case of horizontal gene transfer from Bacteria to Archaea based on phylogenetic profiles, trees, and networks. This means that, finally, the last subfamily of aminoacyl-tRNA synthetases has lost its exceptional position as the sole subfamily that is devoid of horizontal gene transfer. Furthermore, the leucyl-tRNA synthetase phylogenetic tree suggests a dichotomy of the archaeal/eukaryotic-cytosolic and bacterial/eukaryotic-mitochondrial proteins. We argue that the traditional division of life into Prokaryota (non-chimeric) and Eukaryota (chimeric) is favorable compared to Woese's trichotomy into Archaea/Bacteria/Eukaryota.

Transcriptional census of 36 microdissected colorectal cancers yields a gene signature to distinguish UICC II and III.

UICC stage II and III colorectal cancers (CRC) differ fundamentally in prognosis and therapeutic concepts. To analyze differential gene expression between both stages and to establish a relationship between molecular background and clinical presentation, tumor material from 36 unselected consecutive patients presenting with sporadic CRC, 18 UICC stage II and 18 UICC stage III, were laser microdissected to separate epithelial tumor cells. Gene expression levels were measured using U133A Affymetrix gene arrays. Twelve CRC associated signal transduction pathways as well as all 22,000 probe sets were screened for differential gene expression. We identified a signature consisting of 45 probe sets that allowed discrimination between UICC stage II and stage III with a rate of correct classification of about 80%. The most distinctive elements in this signature were the gene GSTP-binding elongation factor (GSPT2) and the transcription factor HOXA9. Differential expression of these genes was confirmed by quantitative real-time polymerase chain reaction (p(HOXA9) = 0.04, p(GSTP2) = 0.02). Despite the reliability of the presented data, there was no substantial differential expression of genes in cancer-related pathways. However, the comparison with recently published data corroborates the 45 gene signature showing structural agreement in the direction of fold changes of gene expression levels for our set of genes chosen to discriminate between both stages.

Nonribosomal peptide synthesis in Schizosaccharomyces pombe and the architectures of ferrichrome-type siderophore synthetases in fungi.

A nonribosomal peptide synthetase (NRPS) in Schizosaccharomyces pombe, which possesses an unusual structure incorporating three adenylation domains, six thiolation domains and six condensation domains, has been shown to produce the cyclohexapeptide siderophore ferrichrome. One of the adenylation domains is truncated and contains a distorted key motif. Substrate-binding specificities of the remaining two domains were assigned by molecular modelling to glycine and to N-acetyl-N-hydroxy-L-ornithine. Hexapeptide siderophore synthetase genes of Magnaporthe grisea and Fusarium graminearum were both identified and analyzed with respect to substrate-binding sites, and the predicted product ferricrocin was identified in each. A comparative analysis of these synthetase systems, including those of the basidiomycete Ustilago maydis, the homobasidiomycete Omphalotus olearius and the ascomycetes Aspergillus nidulans, Aspergillus fumigatus, Fusarium graminearum, Cochliobolus heterostrophus, Neurospora crassa and Aureobasidium pullulans, revealed divergent domain compositions with respect to their number and positioning, although all produce similar products by iterative processes. A phylogenetic analysis of both NRPSs and associated L-N5-ornithine monooxygenases revealed that ferrichrome-type siderophore biosynthesis has coevolved in fungi with varying in trans interactions of NRPS domains.

About the interrelation of evolutionary rate and protein age.

Evolutionary rate and gene age are interrelated when the age of a gene is assessed by the taxonomic distribution in the gene family. This is because homology detection by sequence comparison is depending on sequence similarity. We estimate family specific rates of protein evolution for orthologous families with representatives from man, fugu, fly, and worm. In fact, we observe that younger proteins tend to evolve faster than older ones. We estimate time points of duplication events that gave rise to novel protein functions and show that younger proteins were duplicated more recently than older ones.