Complementary feature selection from alternative splicing events and gene expression for phenotype prediction.

MOTIVATION: A central task of bioinformatics is to develop sensitive and specific means of providing medical prognoses from biomarker patterns. Common methods to predict phenotypes in RNA-Seq datasets utilize machine learning algorithms trained via gene expression. Isoforms, however, generated from alternative splicing, may provide a novel and complementary set of transcripts for phenotype prediction. In contrast to gene expression, the number of isoforms increases significantly due to numerous alternative splicing patterns, resulting in a prioritization problem for many machine learning algorithms. This study identifies the empirically optimal methods of transcript quantification, feature engineering and filtering steps using phenotype prediction accuracy as a metric. At the same time, the complementary nature of gene and isoform data is analyzed and the feasibility of identifying isoforms as biomarker candidates is examined. RESULTS: Isoform features are complementary to gene features, providing non-redundant information and enhanced predictive power when prioritized and filtered. A univariate filtering algorithm, which selects up to the N highest ranking features for phenotype prediction is described and evaluated in this study. An empirical comparison of pipelines for isoform quantification is reported by performing cross-validation prediction tests with datasets from human non-small cell lung cancer (NSCLC) patients, human patients with chronic obstructive pulmonary disease (COPD) and amyotrophic lateral sclerosis (ALS) transgenic mice, each including samples of diseased and non-diseased phenotypes. AVAILABILITY AND IMPLEMENTATION: https://github.com/clabuzze/Phenotype-Prediction-Pipeline.git CONTACT: clabuzze@iastate.edu, antoniom@bc.edu, watsondk@musc.edu, andersonpe2@cofc.edu.

Multivariate models from RNA-Seq SNVs yield candidate molecular targets for biomarker discovery: SNV-DA.

BACKGROUND: It has recently been shown that significant and accurate single nucleotide variants (SNVs) can be reliably called from RNA-Seq data. These may provide another source of features for multivariate predictive modeling of disease phenotype for the prioritization of candidate biomarkers. The continuous nature of SNV allele fraction features allows the concurrent investigation of several genomic phenomena, including allele specific expression, clonal expansion and/or deletion, and copy number variation. RESULTS: The proposed software pipeline and package, SNV Discriminant Analysis (SNV-DA), was applied on two RNA-Seq datasets with varying sample sizes sequenced at different depths: a dataset containing primary tumors from twenty patients with different disease outcomes in lung adenocarcinoma and a larger dataset of primary tumors representing two major breast cancer subtypes, estrogen receptor positive and triple negative. Predictive models were generated using the machine learning algorithm, sparse projections to latent structures discriminant analysis. Training sets composed of RNA-Seq SNV features limited to genomic regions of origin (e.g. exonic or intronic) and/or RNA-editing sites were shown to produce models with accurate predictive performances, were discriminant towards true label groupings, and were able to produce SNV rankings significantly different from than univariate tests. Furthermore, the utility of the proposed methodology is supported by its comparable performance to traditional models as well as the enrichment of selected SNVs located in genes previously associated with cancer and genes showing allele-specific expression. As proof of concept, we highlight the discovery of a previously unannotated intergenic locus that is associated with epigenetic regulatory marks in cancer and whose significant allele-specific expression is correlated with ER+ status; hereafter named ER+ associated hotspot (ERPAHS). CONCLUSION: The use of models from RNA-Seq SNVs to identify and prioritize candidate molecular targets for biomarker discovery is supported by the ability of the proposed method to produce significantly accurate predictive models that are discriminant towards true label groupings. Importantly, the proposed methodology allows investigation of mutations outside of exonic regions and identification of interesting expressed loci not included in traditional gene annotations. An implementation of the proposed methodology is provided that allows the user to specify SNV filtering criteria and cross-validation design during model creation and evaluation.

Hematopoietic stem cell-derived cancer-associated fibroblasts are novel contributors to the pro-tumorigenic microenvironment.

Targeting the tumor microenvironment is critical toward improving the effectiveness of cancer therapeutics. Cancer-associated fibroblasts (CAFs) are one of the most abundant cell types of the tumor microenvironment, playing an important role in tumor progression. Multiple origins for CAFs have been proposed including resident fibroblasts, adipocytes, and bone marrow. Our laboratory previously identified a novel hematopoietic stem cell (HSC) origin for CAFs; however, the functional roles of HSC-derived CAFs (HSC-CAFs) in tumor progression have not yet been examined. To test the hypothesis that HSC-CAFs promote tumor progression through contribution to extracellular matrix (ECM) and paracrine production of pro-angiogenic factors, we developed a method to isolate HSC-CAFs. HSC-CAFs were profiled on the basis of their expression of hematopoietic and fibroblastic markers in two murine tumor models. Profiling revealed production of factors associated with ECM deposition and remodeling. Functional in vivo studies showed that co-injection of HSC-CAFs with tumor cells resulted in increased tumor growth rate and significantly larger tumors than tumor cells alone. Immunohistochemical studies revealed increased blood vessel density with co-injection, demonstrating a role for HSC-CAFs in tumor vascularization. Mechanistic in vitro studies indicated that HSC-CAFs play a role in producing vascular endothelial growth factor A and transforming growth factor-ß1 in endothelial tube formation and patterning. In vitro and in vivo findings suggest that HSC-CAFs are a critical component of the tumor microenvironment and suggest that targeting the novel HSC-CAF may be a promising therapeutic strategy.

FLI1 expression is correlated with breast cancer cellular growth, migration, and invasion and altered gene expression.

ETS factors have been shown to be dysregulated in breast cancer. ETS factors control the expression of genes involved in many biological processes, such as cellular proliferation, differentiation, and apoptosis. FLI1 is an ETS protein aberrantly expressed in retrovirus-induced hematological tumors, but limited attention has been directed towards elucidating the role of FLI1 in epithelial-derived cancers. Using data mining, we show that loss of FLI1 expression is associated with shorter survival and more aggressive phenotypes of breast cancer. Gain and loss of function cellular studies indicate the inhibitory effect of FLI1 expression on cellular growth, migration, and invasion. Using Fli1 mutant mice and both a transgenic murine breast cancer model and an orthotopic injection of syngeneic tumor cells indicates that reduced Fli1 contributes to accelerated tumor growth. Global expression analysis and RNA-Seq data from an invasive human breast cancer cell line with over expression of either FLI1 and another ETS gene, PDEF, shows changes in several cellular pathways associated with cancer, such as the cytokine-cytokine receptor interaction and PI3K-Akt signaling pathways. This study demonstrates a novel role for FLI1 in epithelial cells. In addition, these results reveal that FLI1 down-regulation in breast cancer may promote tumor progression.

Understanding the role of ETS-mediated gene regulation in complex biological processes.

Ets factors are members of one of the largest families of evolutionarily conserved transcription factors, regulating critical functions in normal cell homeostasis, which when perturbed contribute to tumor progression. The well-documented alterations in ETS factor expression and function during cancer progression result in pleiotropic effects manifested by the downstream effect on their target genes. Multiple ETS factors bind to the same regulatory sites present on target genes, suggesting redundant or competitive functions. The anti- and prometastatic signatures obtained by examining specific ETS regulatory networks will significantly improve our ability to accurately predict tumor progression and advance our understanding of gene regulation in cancer. Coordination of multiple ETS gene functions also mediates interactions between tumor and stromal cells and thus contributes to the cancer phenotype. As such, these new insights may provide a novel view of the ETS gene family as well as a focal point for studying the complex biological control involved in tumor progression. One of the goals of molecular biology is to elucidate the mechanisms that contribute to the development and progression of cancer. Such an understanding of the molecular basis of cancer will provide new possibilities for: (1) earlier detection, as well as better diagnosis and staging of disease; (2) detection of minimal residual disease recurrences and evaluation of response to therapy; (3) prevention; and (4) novel treatment strategies. Increased understanding of ETS-regulated biological pathways will directly impact these areas.

Detection of human telomerase reverse transcriptase mRNA in cells obtained by lavage of the pleura is not associated with worse outcome in patients with stage I/II non-small cell lung cancer: results from Cancer and Leukemia Group B 159902.

OBJECTIVE: Previous studies suggest that cytologic analysis of cells obtained by lavage of the pleural surfaces at the time of resection of non-small cell lung cancer can identify patients at risk for recurrence. Because telomerase gene expression has been associated with worse outcome in non-small cell lung cancer, we hypothesized that identification of cells obtained from pleural lavage that express telomerase would identify patients at risk for recurrent disease. METHODS: Patients with presumed non-small cell lung cancer underwent thoracotomy with curative intent. Cells obtained by lavage of the pleural surfaces were analyzed for telomerase catalytic subunit human telomerase reverse transcriptase mRNA expression using reverse transcriptase polymerase chain reaction. RESULTS: A total of 194 patients with stage I/II non-small cell lung cancer had adequate samples, and median follow-up was 60 months (17-91 months). By using Cox models, no statistical differences were found between human telomerase reverse transcriptase-negative and positive patients in disease-free survival (hazard ratio, 1.28; 95% confidence interval, 0.85-1.94; log-rank test, P = .2349) or overall survival (hazard ratio, 1.13; 95% confidence interval, 0.72-1.79; log-rank test, P = .5912) CONCLUSIONS: Detection of human telomerase reverse transcriptase in cells obtained from pleural lavage of patients with stage I/II non-small cell lung cancer does not identify patients at risk for recurrent disease.

Novel immunocompetent murine models representing advanced local and metastatic pancreatic cancer.

BACKGROUND: The development of novel therapeutics for pancreatic cancer has been hindered by a lack of relevant preclinical models. The purpose of this study was to evaluate the clinical relevancy of two pancreatic cancer models using standard-of-care therapeutic agent gemcitabine. MATERIALS AND METHODS: Murine Panc02 cells were injected directly into the spleen or pancreas of C57BL/6 mice to respectively create models of metastatic and locally advanced pancreatic cancer. Beginning 7 d post-Panc02 injection, treated mice received 20 mg/kg gemcitabine i.p. every 3 d. Animals were sacrificed when the untreated mice became moribund and tumor/liver weight used to assess tumor burden. RESULTS: Untreated mice became moribund 22 d after pancreatic Panc02 injection. Gross analysis revealed localized pancreatic tumors weighing 1.063 g. Intrasplenic Panc02 injection produced extensive liver metastasis by d 15 when the untreated mice first became moribund. Liver weights at this time averaged 3.6 g compared with the average non-tumor-bearing weight of 1.23 g. Gemcitabine therapy resulted in a 54% decrease in localized pancreatic tumor weight and 62.5% decrease in metastatic liver weight. Additionally, gemcitabine therapy extended animal survival to 20.5 d compared with 18.0 d average for the untreated mice. CONCLUSIONS: We describe two models depicting both locally advanced and metastatic pancreatic cancer in immunocompetent mice. In efforts to establish baseline therapeutic efficacy, we determined that gemcitabine reduces tumor burden in both models and enhances survival in the metastatic model. These clinically relevant models provide valuable tools to evaluate novel therapeutics in pancreatic cancer.

Prostate-Derived ETS Factor Regulates Epithelial-to-Mesenchymal Transition through Both SLUG-Dependent and Independent Mechanisms.

The 5-year survival rate is very low when breast cancer becomes metastatic. The metastatic process is governed by a network of molecules of which SLUG is known to play a major role as a regulator of epithelial-to-mesenchymal transition (EMT). Prostate-derived ETS factor (PDEF) has been proposed as a tumor suppressor, possibly through inhibition of invasion and metastasis; therefore, understanding the mechanism of PDEF regulation may help to better understand its role in breast cancer progression. This study shows for the first time that the transcription factor SLUG is a direct target of PDEF in breast cancer. We show that the expression of PDEF is able to suppress/dampen EMT through the negative regulation of SLUG. In addition, we show that PDEF is also able to regulate downstream targets of SLUG, namely E-cadherin, in both SLUG-dependent and -independent manners, suggesting a critical role for PDEF in regulating EMT.

Mechanisms and functional consequences of PDEF protein expression loss during prostate cancer progression.

BACKGROUND: Ets is a large family of transcriptional regulators with functions in most biological processes. While the Ets family gene, prostate-derived epithelial factor (PDEF), is expressed in epithelial tissues, PDEF protein expression has been found to be reduced or lost during cancer progression. The goal of this study was to examine the mechanism for and biologic impact of altered PDEF expression in prostate cancer. METHODS: PDEF protein expression of prostate specimens was examined by immunohistochemistry. RNA and protein expression in cell lines were measured by q-PCR and Western blot, respectively. Cellular growth was determined by quantifying viable and apoptotic cells over time. Cell cycle was measured by flow cytometry. Migration and invasion were determined by transwell assays. PDEF promoter occupancy was determined by chromatin immunoprecipitation (ChIP). RESULTS: While normal prostate epithelium expresses PDEF mRNA and protein, tumors show no or decreased PDEF protein expression. Re-expression of PDEF in prostate cancer cells inhibits cell growth. PDEF expression is inversely correlated with survivin, urokinase plasminogen activator (uPA) and slug expression and ChIP studies identify survivin and uPA as direct transcriptional targets of PDEF. This study also shows that PDEF expression is regulated via a functional microRNA-204 (miR-204) binding site within the 3'UTR. Furthermore, we demonstrate the biologic significance of miR-204 expression and that miR-204 is over-expressed in human prostate cancer specimens. CONCLUSIONS: Collectively, the reported studies demonstrate that PDEF is a negative regulator of tumor progression and that the miR-204-PDEF regulatory axis contributes to PDEF protein loss and resultant cancer progression.

Hematopoietic stem cell origin of connective tissues.

Connective tissue consists of "connective tissue proper," which is further divided into loose and dense (fibrous) connective tissues and "specialized connective tissues." Specialized connective tissues consist of blood, adipose tissue, cartilage, and bone. In both loose and dense connective tissues, the principal cellular element is fibroblasts. It has been generally believed that all cellular elements of connective tissue, including fibroblasts, adipocytes, chondrocytes, and bone cells, are generated solely by mesenchymal stem cells. Recently, a number of studies, including those from our laboratory based on transplantation of single hematopoietic stem cells, strongly suggested a hematopoietic stem cell origin of these adult mesenchymal tissues. This review summarizes the experimental evidence for this new paradigm and discusses its translational implications.

Hematopoietic stem cell origin of mesenchymal cells: opportunity for novel therapeutic approaches.

There has been a general belief that there are two types of adult stem cells, i.e., hematopoietic stem cells (HSCs) and mesenchymal stem cells (MSCs), each with distinctly different functions. According to this dogma, HSCs produce blood cells, while MSCs are thought to generate a number of non-hematopoietic cells including fibroblasts, adipocytes, chondrocytes and bone cells. Recently, a number of studies, including those in our laboratory based on single HSC transplantation, blurred the clear distinction between HSCs and MSCs and strongly suggested an HSC origin of the adult mesenchymal tissues. This review summarizes the experimental evidence for this new paradigm and the literature pointing out the vagary in the stem cell nature of MSCs. The concept of the HSC origin of mesenchymal cells will have many immediate and long-term impacts on the therapies of diseases and injuries of the connective tissues.

Hematopoietic stem cell origin of adipocytes.

OBJECTIVE: It has generally been believed that adipocytes are derived from mesenchymal stem cells via fibroblasts. We recently reported that fibroblasts/myofibroblasts in a number of tissues and organs are derived from hematopoietic stem cells (HSCs). In the present study, we tested the hypothesis that HSCs also give rise to adipocytes. MATERIALS AND METHODS: Using transplantation of a single enhanced green fluorescent protein-positive (EGFP(+)) HSC and primary culture, we examined generation of adipocytes from HSCs. RESULTS: Adipose tissues from clonally engrafted mice showed EGFP(+) adipocytes that stained positive for leptin, perilipin, and fatty acid binding protein 4. A diet containing rosiglitazone, a peroxisome proliferator-activated receptor-gamma agonist, significantly enhanced the number of EGFP(+) adipocytes. When EGFP(+) bone marrow cells from clonally engrafted mice were cultured under adipogenic conditions, all of the cultured cells stained positive with Oil Red O and Sudan Black B and exhibited the presence of abundant mRNA for adipocyte markers. Finally, clonal culture- and sorting-based studies of Mac-1 expression of hematopoietic progenitors suggested that adipocytes are derived from HSCs via progenitors for monocytes/macrophages. CONCLUSION: Together, these studies clarify the current controversy regarding the ability of HSCs to give rise to adipocytes. Furthermore, our primary culture method that generates adipocytes from uncommitted hematopoietic cells should contribute to the studies of the mechanisms of early adipocytic differentiation and may lead to development of therapeutic solutions for many general obesity issues.

CaSm (LSm-1) overexpression in lung cancer and mesothelioma is required for transformed phenotypes.

CaSm (cancer-associated Sm-like) was originally identified based on elevated expression in pancreatic cancer and in several cancer-derived cell lines. It encodes a 133-amino acid protein that contains two Sm motifs found in the common snRNP proteins and the LSm (like-Sm) family of proteins. Lung tumors and mesotheliomas express high levels of CaSm mRNA and protein compared with adjacent nontumor and normal lung tissue, measured by immunohistochemistry, qRT-PCR, and Western blot analyses. In addition, several human lung cancer- and mesothelioma-derived cell lines have elevated CaSm expression. Two cell lines, transfected with and expressing antisense CaSm RNA, demonstrate altered transformed phenotypes, reducing their ability to form colonies in soft agar and tumors in SCID mice. Furthermore, RNAi-mediated reduction of CaSm RNA and protein is associated with inhibition of cellular growth. These data support the model that elevated CaSm expression in epithelial tissue contributes to the transformed state. Cell lines expressing exogenous CaSm also exhibit transformed characteristics, including increased anchorage-independent colony formation and tumor growth. Thus, the results of loss of function and gain of function studies presented both indicate that CaSm functions as an oncogene in the promotion of cellular transformation and cancer progression.

CaSm-mediated cellular transformation is associated with altered gene expression and messenger RNA stability.

CaSm (cancer-associated Sm-like) was originally identified based on elevated expression in pancreatic cancer and in several cancer-derived cell lines. CaSm encodes a 133 amino acid protein that contains two Sm motifs found in the common small nuclear RNA proteins and the LSm (like-Sm) family of proteins. Compared with normal human prostate tissue and primary prostate epithelial cells, some primary prostate tumors and prostate cancer-derived cell lines have elevated CaSm expression. Expression of antisense CaSm RNA in DU145 cells results in reduced CaSm protein levels and less transformed phenotype, measured by anchorage-independent growth in vitro and tumor formation in severe combined immunodeficient mice in vivo. Additional data shows that adenoviral delivery of antisense CaSm inhibits the growth of prostate cancer cell lines by altering cell cycle progression, and is associated with reduced expression of cyclin B1 and CDK1 proteins. Consistent with failure of antisense-treated cells to enter mitosis, microarray analysis identified altered expression of NEK2 and nucleophosmin/B23. Although the mechanisms by which CaSm contributes to neoplastic transformation and cellular proliferation are unknown, it has been shown that the yeast homologue (spb8/LSm1) of CaSm is required for 5' to 3' degradation of specific mRNAs. We provide data consistent with a similar role for CaSm in human cells, supporting the hypothesis that elevated CaSm expression observed in cancer leads to destabilization of multiple gene transcripts, contributing to the mutator phenotype of cancer cells.

Establishing a murine pancreatic cancer CaSm model: up-regulation of CaSm is required for the transformed phenotype of murine pancreatic adenocarcinoma.

We have recently shown that the cancer-associated Sm-like protein (CaSm) is overexpressed in human pancreatic adenocarcinoma (PC). However, the role of CaSm in the process of neoplastic transformation remains unclear. To define further the role of CaSm in PC transformation, we have established a murine model based on the murine pancreatic cancer cell lines Panc02 and Panc03. CaSm is overexpressed in the aggressive Panc02 cells and expressed at much lower levels in the more indolent Panc03 cells. Up-regulation of CaSm in Panc03 cells increased in vitro proliferation and anchorage-independent growth and promoted subcutaneous tumor establishment and growth in syngeneic mice. Conversely, adenoviral down-regulation of CaSm in Panc02 led to significant inhibition of cellular proliferation and anchorage-independent growth in vitro and complete abolition of tumor growth and metastasis in vivo. Up-regulation of CaSm in NIH3T3 resulted in loss of contact inhibition and increased soft agar colony formation in vitro. The requirement for CaSm overexpression for neoplastic transformation confirms the concept that CaSm is a critical oncogene and potential target for molecular intervention. Furthermore, establishment of the murine clinically relevant model of pancreatic metastases provides a framework for the generation of preclinical data to support the development of novel molecular therapies targeting CaSm.

Differential mechanisms and development of leptin resistance in A/J versus C57BL/6J mice during diet-induced obesity.

Changes in the biological efficacy of leptin were evaluated in obesity-resistant (A/J) and obesity-prone (C57BL/6J) mice at weaning and after consuming a high-fat (HF) diet for 4 and 8 wk. There was no evidence of leptin resistance in either strain at the start of the study, but after 4 and 8 wk on the HF diet, C57BL/6J mice became unresponsive to ip leptin. C57BL/6J mice responded to intracerebroventricular leptin at these time points but developed peripheral resistance to sympathetic stimulation of retroperitoneal white adipose tissue. In contrast, intracerebroventricular leptin was fully effective in A/J mice, reproducing the complete profile of responses observed in weanling mice. A/J mice were also partially responsive to ip leptin at both time points, increasing uncoupling protein 1 mRNA expression in brown adipose tissue and decreasing leptin mRNA in white adipose tissue. The findings indicate that retention of leptin responsiveness is an important component of the ability of A/J mice to mount a robust adaptive thermogenic response and resist diet-induced obesity.

Adaptive changes in adipocyte gene expression differ in AKR/J and SWR/J mice during diet-induced obesity.

Obesity-prone (AKR/J) and obesity-resistant (SWR/J) mice were weaned onto low (LF) or high fat (HF) diets to identify adaptive changes in adipocyte gene expression that are associated with differences between the strains in fat deposition. Food consumption was monitored at weekly intervals and all mice were evaluated after consuming their respective diets for 4 wk for analysis of mRNA levels of selected metabolic genes. Despite similar food consumption, body weight and fat deposition were significantly greater in AKR/J than in SWR/J mice, and this difference was greatly accentuated by the HF diet. The HF diet produced distinct differences between strains in gene expression patterns among fat depots. In AKR/J mice, UCP1 mRNA was decreased 10-fold in interscapular brown adipose tissue (BAT) and four- to fivefold in retroperitoneal and inguinal white adipose tissue (WAT). The HF diet also decreased PGC-1 and beta(3)-adrenergic receptor mRNA by two- and ninefold in BAT from AKR/J mice. In contrast, the HF diet either increased uncoupling protein (UCP)1 in BAT or had no effect on expression of these genes in adipose tissues from SWR/J mice. UCP2 mRNA was fourfold higher in WAT from AKR/J compared with SWR/J mice and increased by an additional twofold in WAT from AKR/J mice fed the HF diet. UCP2 was unaffected by diet in SWR/J mice. These studies show that the diet-induced obesity of AKR/J mice is characterized by increased metabolic efficiency and is associated with changes in adipocyte gene expression that limit the adaptive thermogenic response to increased energy density.