Raw Gene Expression Research Articles

Genome-wide transcriptional profiling data from chronic cutaneous lupus erythematosus (CCLE) peripheral blood.

The disease Lupus erythematosus (LE), exhibits a variety of clinical manifestations with potentially wide-ranging multi-organ damage to joints, tendons, kidney, lung, heart, blood vessels, central nervous system and skin [1,2] Systemic changes are likely to trigger organ specific manifestation of the disease. Here, we provide the data examined to address the gap in knowledge regarding causes and mechanisms contributing to the autoimmune attack on skin in chronic cutaneous lupus erythematosus (CCLE). The raw gene expression data files (CEL files) are provided with this article [3].

Conditional Density Estimation in Measurement Error Problems.

This paper is motivated by a wide range of background correction problems in gene array data analysis, where the raw gene expression intensities are measured with error. Estimating a conditional density function from the contaminated expression data is a key aspect of statistical inference and visualization in these studies. We propose re-weighted deconvolution kernel methods to estimate the conditional density function in an additive error model, when the error distribution is known as well as when it is unknown. Theoretical properties of the proposed estimators are investigated with respect to the mean absolute error from a "double asymptotic" view. Practical rules are developed for the selection of smoothing-parameters. Simulated examples and an application to an Illumina bead microarray study are presented to illustrate the viability of the methods.

Multi-tissue Analysis of Co-expression Networks by Higher-Order Generalized Singular Value Decomposition Identifies Functionally Coherent Transcriptional Modules

Recent high-throughput efforts such as ENCODE have generated a large body of genome-scale transcriptional data in multiple conditions (e.g., cell-types and disease states). Leveraging these data is especially important for network-based approaches to human disease, for instance to identify coherent transcriptional modules (subnetworks) that can inform functional disease mechanisms and pathological pathways. Yet, genome-scale network analysis across conditions is significantly hampered by the paucity of robust and computationally-efficient methods. Building on the Higher-Order Generalized Singular Value Decomposition, we introduce a new algorithmic approach for efficient, parameter-free and reproducible identification of network-modules simultaneously across multiple conditions. Our method can accommodate weighted (and unweighted) networks of any size and can similarly use co-expression or raw gene expression input data, without hinging upon the definition and stability of the correlation used to assess gene co-expression. In simulation studies, we demonstrated distinctive advantages of our method over existing methods, which was able to recover accurately both common and condition-specific network-modules without entailing ad-hoc input parameters as required by other approaches. We applied our method to genome-scale and multi-tissue transcriptomic datasets from rats (microarray-based) and humans (mRNA-sequencing-based) and identified several common and tissue-specific subnetworks with functional significance, which were not detected by other methods. In humans we recapitulated the crosstalk between cell-cycle progression and cell-extracellular matrix interactions processes in ventricular zones during neocortex expansion and further, we uncovered pathways related to development of later cognitive functions in the cortical plate of the developing brain which were previously unappreciated. Analyses of seven rat tissues identified a multi-tissue subnetwork of co-expressed heat shock protein (Hsp) and cardiomyopathy genes (Bag3, Cryab, Kras, Emd, Plec), which was significantly replicated using separate failing heart and liver gene expression datasets in humans, thus revealing a conserved functional role for Hsp genes in cardiovascular disease.

Human Adipose Tissue Macrophages Display Activation of Cancer-related Pathways

Obesity is associated with a significantly increased risk for cancer suggesting that adipose tissue dysfunctions might play a crucial role therein. Macrophages play important roles in adipose tissue as well as in cancers. Here, we studied whether human adipose tissue macrophages (ATM) modulate cancer cell function. Therefore, ATM were isolated and compared with monocyte-derived macrophages (MDM) from the same obese patients. ATM, but not MDM, were found to secrete factors inducing inflammation and lipid accumulation in human T47D and HT-29 cancer cells. Gene expression profile comparison of ATM and MDM revealed overexpression of functional clusters, such as cytokine-cytokine receptor interaction (especially CXC-chemokine) signaling as well as cancer-related pathways, in ATM. Comparison with gene expression profiles of human tumor-associated macrophages showed that ATM, but not MDM resemble tumor-associated macrophages. Indirect co-culture experiments demonstrated that factors secreted by preadipocytes, but not mature adipocytes, confer an ATM-like phenotype to MDM. Finally, the concentrations of ATM-secreted factors related to cancer are elevated in serum of obese subjects. In conclusion, ATM may thus modulate the cancer cell phenotype.

EXP-PAC: Providing comparative analysis and storage of next generation gene expression data

Microarrays and more recently RNA sequencing has led to an increase in available gene expression data. How to manage and store this data is becoming a key issue. In response we have developed EXP-PAC, a web based software package for storage, management and analysis of gene expression and sequence data. Unique to this package is SQL based querying of gene expression data sets, distributed normalization of raw gene expression data and analysis of gene expression data across experiments and species. This package has been populated with lactation data in the international milk genomic consortium web portal (http://milkgenomics.org/). Source code is also available which can be hosted on a Windows, Linux or Mac APACHE server connected to a private or public network (http://mamsap.it.deakin.edu.au/~pcc/Release/EXP_PAC.html)

Abstract 2479: TP53 status as a marker of recurrence and survival in choroid plexus carcinomas

Abstract Choroid plexus carcinoma (CPC) is a rare intraventricular tumor more frequently observed in children, accounting for ∼20% of brain tumors in children under 2 yrs of age. This malignant neoplasm is categorized as a grade III tumor by the WHO, while benign choroid plexus papilloma (CPP) and atypical choroid plexus papilloma (aCPP) are categorized as grade I and II, respectively. More than 50% of CPC's carry a mutation in TP53. The role of p53 in the malignant progression of this tumor is not well understood, yet this knowledge is crucial to optimize patient treatment, increase survival and reduce long-term sequelae. The current treatment approach for CPC includes surgical resection and combination chemo- and radiation therapy. However, treatment response is highly variable; tumor progression and relapse is observed in ∼60% of patients, and most survivors experience significant long-term deficits. In this study, we sought to determine whether TP53 status could be used as a reliable marker to predict tumor aggressiveness and response to treatment in CPC patients. Thirty CPC, 26 CPP and 8 aCPP tumor samples for which TP53 status was available were selected from the SickKids Cancer Genetics tissue bank. Sequencing analysis revealed somatic TP53 in 60% of CPCs. These mutations were significantly less frequent in CPPs (12%, p=1.84x10−4). All aCPPs were wildtype (wt) for TP53. Relapse-free survival was only 12% in TP53 mut CPC patients compared to 40% for TP53 wt patients. Only 1 CPP patient relapsed 9 months after initial diagnosis. Five-year overall survival for TP53 mut CPC patients was significantly lower than that of TP53 wt patients (39% vs. 80%, respectively, p=0.04). Overall survival for 18 CPP patients for which we had 5-year survival data was 94%, and 100% for aCPP patients. Allele-specific copy number (ASCN) and FISH analysis, revealed extensive chromosomal imbalances in CPC tumors with TP53 mutations. TP53 wt CPCs had fewer chromosomal imbalances or none at all, very similar to CPPs. Gene expression data was used to examine the unsupervised clustering of our tumors. As expected, CPCs and CPPs cluster independently according to raw gene expression intensity, with TP53 wt CPCs segregating from the CPC mutant samples and clustering more closely to CPPs. Using an integrative analysis, we were able to identify unique behaviors and molecular characteristics that distinguish TP53 wt from TP53 mut CPCs. The clinical outcomes observed, supported by the genomic and transcriptomic data analyzed, suggest that TP53 mut and wt CPCs indeed form two distinct subgroups. This molecular genotyping approach may be valuable in stratifying patients to modulated therapy protocols in order to diminish treatment-related toxicity, and improve disease response and survival. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr 2479. doi:1538-7445.AM2012-2479

Abstract 5092: Localized and disseminated histiocytic sarcoma complex: A canine model

Abstract Canine histiocytic sarcoma is an aggressive and uniformly-fatal round-cell neoplasm. The localized, peri-articular form of the disease (PAHS) appears to have statistically significant prolonged survival, when compared to the non - peri-articular form of the disease (non-PAHS). Knowledge of the molecular basis for the aggressiveness of the canine histiocytic sarcoma is foundational for developing effective treatments. The aim of this study was to characterize the molecular expression profiles of the different sub-types of the disease. Canine patients, with spontaneously-arising histiocytic sarcoma, were recruited for the study. Tumor tissue samples and peripheral blood were obtained and used for primary cell line development and gene expression analysis. Eight PAHS and 8 non-PAHS tumor samples were selected. RNA extraction, amplification and labeling were performed using standard techniques. Raw gene expression values were obtained using the Affymetrix GeneChip Canine Genome 2.0 Array. Background correction, normalization, PM correction, summarization, and expression analysis were performed using R/affy/limma. Gene set analysis was performed using BRB-ArrayTools, and the Molecular Signatures Database (MSigDB), to investigate gene ontology groups of genes whose expression was differentially regulated. Primary cell lines from tumor biopsies and peripheral blood samples were developed and characterized based on cellular morphology, growth characteristics, and immunophenotyping. In our gene expression array experiments, we identified specific gene transcripts that are more likely to represent true and biologically meaningful differences in expression levels between PAHS and non-PAHS. More specifically, metallopeptidase genes, and homeobox domain gene expression levels were enriched in the PAHS group, while ATP-binding cassette genes, and interleukin 1 receptor signaling gene expression levels were enriched in the non-PAHS group. Our results suggest presence of gene expression signatures that can provide insight in the diverse biologic behavior of the disease and highlight canine spontaneous tumors in understanding tumor biology. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr 5092. doi:1538-7445.AM2012-5092

Quality process in genomic pathology

The fundamental basis of quality control was pioneered by Shewhart in the 1920s, and developed by Demming during the 1940 s. Although these principles were developed in manufacturing, many pathology laboratories have recognised their importance. Whether in manufacturing or in pathology the basic principles remain the same: identify the sources of variation in the finished product, define measures of product quality at each stage in processing, routinely record those quality measures, and develop statistical decision rules to identify when a process is out of control—that is when it is delivering an unacceptable proportion of low quality results. The more complex an assay, the greater need there is for formal quality control processes. Few assays are as complex as those implemented using modern genomic platforms. Yet these assays produce vast quantities of data, which make comprehensive monitoring and the development of appropriate decision rules challenging. We consider a case study involving a multivariate index assay based on gene expression. Raw gene expression data were produced by a pathology laboratory as a technical service to the diagnostics company. The diagnostic company then provided interpretation, through computation of the multivariate index assay. The diagnostic company's quality processes revealed a substantial deterioration of the quality of raw results, and the service laboratory was found to have inadequate controls in place. We provide details of the appropriate quality metrics and decision rules, but the identity of both the diagnostics company and service laboratory are withheld.

Genomic Heterogeneity in B-Cell Malignancies,

Abstract 3465 Background:Chronic lymphocytic leukemia (CLL), aggressive B-cell non-Hodgkin lymphomas (NHL), and multiple myeloma (MM) are B-cell malignancies that display biological and clinical heterogeneity. Current investigations into the genetics and biology of these related disorders are using next generation whole genome or exome sequencing. The relatively high cost of these techniques has driven an experimental design in which a small group of samples are initially studied, specific genetic lesions are identified, and then larger cohorts are evaluated for those specific aberrations. Given the biological heterogeneity that is found in each of these disorders, such an approach could skew the direction of research towards results found in a small subset of patients. To determine the extent of genomic heterogeneity within and similarities between CLL, NHL, and MM, and their biologic and clinical relevance, we evaluated publicly available gene expression and single nucleotide polymorphism (SNP) array data from the NCBI Gene Expression Omnibus. Methods:We analyzed 893, 881, and 1744 unique gene expression data files that represent CLL, NHL, and MM, respectively. The gene expression data files represented 15, 11, and 10 distinct data sets, respectively. Prognostic, clinical outcome, and copy number variation data were available for a subset of the samples from each malignancy. Gene expression data were initially normalized using RMA and MAS5 algorithms and batch effect was eliminated using Bayesian Factor Regression Modeling. SNP array data were normalized using Chromosome Copy Number Analysis Tool and amplifications and deletions were identified with circular binary segmentation. Analyses were carried out using Bioconductor packages and the statistical environment R. Results:After elimination of batch effect, we evaluated the data using random subsampling and unsupervised hierarchical clustering to determine the lowest number of samples required to capture genomic heterogeneity. For CLL and NHL, there was no plateau reached for the number of groups defined by hierarchical clustering up through the total number of samples, indicating that a larger number of samples than available in this study are needed to fully document biological and genomic variability. For MM, there was a plateau reached at approximately 1200 samples. We then used unsupervised hierarchical clustering of the entire dataset for each malignancy to define groups of CLL, NHL, and MM based on their raw gene expression data. To evaluate the biological meaning of the groups defined by this process, we used tools such as Gene Set Enrichment Analysis (GSEA) and oncogenic pathway predictions (ScoreSignatures). Groups within each malignancy that were defined using raw gene expression data had differences in biological pathways involving receptor signaling, cell cycle, and stem cell properties. Notably, similarities in biological annotation were seen between groups that represented the different malignancies. Although prognostic data was not available for all the datasets, there appeared to be no differences in clinical prognostic markers between the genomic-defined groups. However, there were statistically significant differences in molecular prognostic data between these groups. In addition, specific regions of DNA copy number variation were enriched within the different genomic-defined groups. Together, these data highlight the biologic distinctions between groups that are defined by raw gene expression data. For datasets in which clinical outcome data were available, we found that genomic-defined groups had different outcomes such as time to first therapy or overall survival. However, the groups did not appear to predict response to chemotherapy or chemo-immunotherapy. Conclusions:CLL, NHL, and MM are heterogeneous malignancies, and very large numbers of patients must be studied to fully capture the genomic and biologic diversity that is present. Despite this limitation, evaluation of existing data reveals subgroups of these disorders are defined by their underlying biology, demonstrate overlap in biological processes, and are clinically relevant. These results have implications on future “omics” related research. Disclosures:No relevant conflicts of interest to declare.

Who shares? Who doesn't? Factors associated with openly archiving raw research data.

Many initiatives encourage investigators to share their raw datasets in hopes of increasing research efficiency and quality. Despite these investments of time and money, we do not have a firm grasp of who openly shares raw research data, who doesn't, and which initiatives are correlated with high rates of data sharing. In this analysis I use bibliometric methods to identify patterns in the frequency with which investigators openly archive their raw gene expression microarray datasets after study publication.Automated methods identified 11,603 articles published between 2000 and 2009 that describe the creation of gene expression microarray data. Associated datasets in best-practice repositories were found for 25% of these articles, increasing from less than 5% in 2001 to 30%–35% in 2007–2009. Accounting for sensitivity of the automated methods, approximately 45% of recent gene expression studies made their data publicly available.First-order factor analysis on 124 diverse bibliometric attributes of the data creation articles revealed 15 factors describing authorship, funding, institution, publication, and domain environments. In multivariate regression, authors were most likely to share data if they had prior experience sharing or reusing data, if their study was published in an open access journal or a journal with a relatively strong data sharing policy, or if the study was funded by a large number of NIH grants. Authors of studies on cancer and human subjects were least likely to make their datasets available.These results suggest research data sharing levels are still low and increasing only slowly, and data is least available in areas where it could make the biggest impact. Let's learn from those with high rates of sharing to embrace the full potential of our research output.

Abstract 4310: Gene expression analysis of axial and appendicular osteosarcoma in the canine model

Abstract Spontaneously occurring canine osteosarcoma (OSA) is considered an excellent model of the human disease. Canine axial and appendicular osteosarcoma appear histopathologically identical. The clinical course of the two forms, however, seems to diverge. Following surgery, the axial form may be associated with a lower metastatic rate and longer survival than the appendicular form. Identifying gene expression profiles associated with decreased metastatic potential may be crucial for the development of targeted therapies for the human disease. We hypothesized that there are distinct gene expression patterns for canine axial and appendicular OSA. Six axial osteosarcoma tumor samples and six age, breed, and weight matched appendicular osteosarcoma clinical tumor samples were selected from the Center for Comparative Oncology, Michigan State University tissue bank. RNA extraction, amplification and labeling were performed using standard techniques. Raw gene expression values were obtained using the Affymetrix GeneChip Canine Genome 2.0 Array. Background correction, normalization, PM correction, summarization, and expression analysis were performed using the R software with the affy/limma/sva packages. Gene set enrichment analysis was performed using GSEA software. All RNA samples passed quality control and were included in the data set. The expression analysis took into account RNA degradation, which was similar in all tumor samples. A total of 184 genes were differentially expressed by >16-fold, between the two groups. Surrogate variable analysis did not result in any statistically significant differences in gene expression. Gene set enrichment analysis resulted in 9 gene sets enriched in axial, and 4 gene sets enriched in appendicular osteosarcoma samples at FDR < 25%. Gene ranking was similar between the two methods of analysis, with MMP1, ATP6AP1, SLC6A8, and dysadherin expression increased in axial tumor samples, while ICAM2 expression increased in appendicular samples. Expression analysis indicates no statistically significant differences in gene expression between the axial and appendicular OSA. Gene set enrichment analysis indicates potential differences in gene set enrichment between the two types of osteosarcoma. Further quantitative RT-PCR analysis is required to confirm our results. These data will guide hypothesis-driven translational studies, which may have direct clinical implications on the treatment of osteosarcoma. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 102nd Annual Meeting of the American Association for Cancer Research; 2011 Apr 2-6; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2011;71(8 Suppl):Abstract nr 4310. doi:10.1158/1538-7445.AM2011-4310

A novel Wavelet-Based Clustering Algorithm for Analysis of Gene Expression Patterns

Most computer clustering programs available today use global de-noising algorithms to filter raw gene expression pattern data. However, many gene expression time series data contain interval-dependent noise where the second moment statistics of the noise are non-stationary. To address this issue we developed a new wavelet-based algorithm (Wave-SOM) that uses a localized filtering method (wavelets) to remove noise from the data while preserving local time events in the gene expression patterns. We employed a discrete Hilbert transform thresholding technique to compare the size of the signal component relative to the noise level at each wavelet transform level by creating a complex-valued analytic vector from which an amplitude vector was defined. Using various wavelet transformations, raw data are first de-noised by decomposing the time-series into low and high frequency wavelet coefficients. Following thresholding, the coefficients are fed as an input vector into a two-dimensional Self-Organizing-Map clustering algorithm. Transformed data are then clustered by minimizing the Euclidean (L2) distance between their corresponding fluctuation patterns. A multi-resolution analysis by Wave-SOM of expression data from the yeast Saccharomyces cerevisiae, exposed to oxidative stress and glucose-limited growth identified 29 genes with correlated expression patterns that were mapped into 5 different nodes. This ordered clustering of yeast genes by Wave-SOM illustrates the fact that the same set of genes (encoding ribosomal proteins) can be regulated by two different environmental stresses, oxidative stress and starvation. Using an adjusted Rand index measure to cluster expression patterns of yeast’ cell-cycle genes as test data sets, our algorithm outperformed the Cluster 3.0, MCLUST, CurveSOM , SSClust and GENECLUSTER clustering algorithms.

Transcriptome Analysis of Differentiating Erythroid Progenitors in Refractory Anemia with Ringed Sideroblasts

AbstractAbstract 1864Refractory anaemia with ring sideroblasts (RARS) represents one subtype of the myelodysplastic syndrome characterized by apoptosis of erythroid precursors and accumulation of aberrant mitochondrial ferritin. Gene expression profiling (GEP) has revealed significant dysregulation of genes involved in iron transport and mitochondrial and erythroid function (Nikpour et al, 2010) but no specific mutations have been identified. Generally, dysregulation increases with forced differentiation indicating alterations of underlying transcription factors. To further explore the molecular mechanisms in RARS, we examined the transcriptional profile associated with early erythroid maturation in NBM and RARS using RNA-Seq. RNA sequencing (RNA-Seq) is a novel method to analyze the full transcriptional activity of a cell or tissue. Expression levels of specific genes, differential splicing, allele-specific expression of transcripts can be accurately determined by RNA-Seq experiments to address many biological-related issues.Bone marrow aspirates were collected from a patient and a control and subjected to CD34+ separation. Cells were cultured for 4 days to allow for erythroid maturation. cDNA libraries were prepared from RNA extracted from these two time points (0 and 4 days), and thereafter sequenced on Life Technology's next generation sequencing platform SOLiD. Approximately 80 million reads were obtained for each library (see Table 1). Two sets of analyses were made; one blinded for position (PU) to allow for an unbiased analysis of data and one (MN) comparing position profiles with raised previous GEP findings.Table 1:Read mapping summarySampleReads(%)unmapped(%)mapped(%)unique(%)multi(%)Control (CD34+)80955215100.06071678075.02023843525.01679855620.834398794.2Control (D4)80054221100.05748309771.82257112428.21802121522.545499095.7RARS (CD34+)79567250100.06010721175.51946003924.51565008019.738099594.8RARS (D4)73764107100.05606428776.01769982024.01404054419.036592765.0Reads were mapped to the human genome reference sequence (NCBI build version GRCh37) using the bowtie alignment program. To minimize expression level bias through the misclassification of transcripts, only uniquely mapping reads were retained for downstream analysis. Between 14 and 16 million uniquely mapping reads were obtained for each library.Gene models were defined on the genome based on annotations in the ensemble genome database release 58. Reads were associated with a gene if it overlapped any part of a gene, including introns. Furthermore, overlapping genes were removed, ensuring unique read-gene association. 38926 gene models were used in the analysis. Raw gene expression counts for each gene model were obtained as the number of reads that overlapped that gene model. Between 7 and 10 million reads were retained for further analysis.Differential gene expression analysis was done using the bioconductor package DESeq. We compared samples pairwise using a negative binomial statistical model to assess significant differential expression. We corrected for multiple testing using a false discovery rate of 0.1. The number of differentially expressed genes ranged from 10 (comparison Control D4-RARS D4) to 294 (comparison Control CD34+-RARS CD34+). Due to the lack of replicates, variance estimates are uncertain, leading to a lower number of inferred differentially expressed genes. Based on correlation of gene expression and the number of differentially expressed genes, samples could be clustered into two pairs, Control CD34+-RARS CD34+ and Control D4-RARS D4.In order to infer potential functional differences between the samples, we analyzed the function of the differentially expressed genes and performed gene category overrepresentation analysis, using the gene ontology classification system. Interestingly, several non-coding RNAs involved in eg miRNA processing were significantly down-regulated in both RARS positions, compared to NBM. We also identified dramatic dysregulation of two putative zinc finger transcription factors during erythroid differentiation.We conclude that high-quality transcriptome analysis at different time points during erythroid maturation is able to add significant new information to conventional gene expression profiling, which will lead to a more comprehensive understanding of the molecular pathogenesis of RARS. Disclosures:No relevant conflicts of interest to declare.

Who shares? Who doesn't? Bibliometric factors associated with open archiving of biomedical datasets

Many initiatives encourage investigators to share their raw research datasets in pursuit of increased research quality and efficiency. Despite these investments of time and money, we do not yet understand the impact of these initiatives. In this study, I use bibliometric methods to understand the prevalence and patterns with which investigators publicly share their raw gene expression microarray datasets after study publication. Automated methods were used to identify 11,603 published studies that created gene expression microarray data. At least 25% of these studies have datasets in one of the two predominant public databases for microarray data, increasing from 5% in 2001 to 35% in 2009. Fifteen factors that described authorship, funding, institution, publication, and domain environments were derived from 124 article attributes. Most factors associated with the prevalence of data sharing (p<0.01). In particular, publishing in a journal with a relatively strong data sharing policy, receiving funding from many NIH grants, publishing in an open access journal, and having prior experience sharing gene expression data were associated with the highest data sharing rates. In contrast, increased first author age and experience, having no experience reusing data, and studying cancer and human subjects were associated with the lowest data sharing rates. In second-order factor analysis, previously sharing gene expression microarray data was most positively associated with high data sharing rates, whereas publishing a study on cancer or human subjects was strongly associated with a negative probability of data sharing. I hope these methods and results will contribute to a deeper understanding of data sharing behavior and eventually more effective data sharing initiatives.

Genome-wide expression profiling reveals new candidate genes associated with osteoarthritis

Although the extracellular matrix (ECM) is the functional element in articular cartilage and its degradation is central in the pathogenetic process in osteoarthritis (OA), increasing the knowledge about the cellular OA phenotype is essential. The aim of this study is therefore to provide a more complete picture of the cellular and molecular alterations detected in OA cartilage. Human articular cartilage biopsies were collected from donors with macroscopical and microscopical signs of OA as well as donors with no previous history of OA and with microscopically intact cartilage. RNA was isolated from the biopsies and subjected to whole genome microarray analysis. Important results from the microarray analysis were verified using real-time PCR and immunohistochemistry. Our results reveal several new candidate genes not previously associated with OA to display significantly higher expression in OA cartilage than in normal donor cartilage, including genes involved in bone formation (CLEC3B, CDH11, GPNMB, CLEC3A, CHST11, MSX1, MSX2) and genes encoding collagens (COL13A1, COL14A1, COL15A1, COL8A2). This study is the first to report a comprehensive gene expression analysis of human OA cartilage compared to control cartilage from donors lacking macroscopical and microscopical signs of OA using recently developed microarrays containing the whole human genome. Our results could broadly confirm previously published data on many characteristic features of OA as well as adding a panel of genes to the list of genes known to be differentially expressed in OA. Elucidation of the phenotypical alterations occurring in OA chondrocytes is important for the development of effective treatments for OA.

Genome Holography: Deciphering Function-Form Motifs from Gene Expression Data

BackgroundDNA chips allow simultaneous measurements of genome-wide response of thousands of genes, i.e. system level monitoring of the gene-network activity. Advanced analysis methods have been developed to extract meaningful information from the vast amount of raw gene-expression data obtained from the microarray measurements. These methods usually aimed to distinguish between groups of subjects (e.g., cancer patients vs. healthy subjects) or identifying marker genes that help to distinguish between those groups. We assumed that motifs related to the internal structure of operons and gene-networks regulation are also embedded in microarray and can be deciphered by using proper analysis.Methodology/Principal FindingsThe analysis presented here is based on investigating the gene-gene correlations. We analyze a database of gene expression of Bacillus subtilis exposed to sub-lethal levels of 37 different antibiotics. Using unsupervised analysis (dendrogram) of the matrix of normalized gene-gene correlations, we identified the operons as they form distinct clusters of genes in the sorted correlation matrix. Applying dimension-reduction algorithm (Principal Component Analysis, PCA) to the matrices of normalized correlations reveals functional motifs. The genes are placed in a reduced 3-dimensional space of the three leading PCA eigen-vectors according to their corresponding eigen-values. We found that the organization of the genes in the reduced PCA space recovers motifs of the operon internal structure, such as the order of the genes along the genome, gene separation by non-coding segments, and translational start and end regions. In addition to the intra-operon structure, it is also possible to predict inter-operon relationships, operons sharing functional regulation factors, and more. In particular, we demonstrate the above in the context of the competence and sporulation pathways.Conclusions/SignificanceWe demonstrated that by analyzing gene-gene correlation from gene-expression data it is possible to identify operons and to predict unknown internal structure of operons and gene-networks regulation.

Oncogene-Induced Senescence Relayed by an Interleukin-Dependent Inflammatory Network

Oncogene-induced cellular senescence (OIS) is emerging as a potent cancer-protective response to oncogenic events, serving to eliminate early neoplastic cells from the proliferative pool. Using combined genetic and bioinformatic analysis, we find that OIS is linked specifically to the activation of an inflammatory transcriptome. Induced genes included the pleiotropic cytokine interleukin-6 (IL-6), which upon secretion by senescent cells acted mitogenically in a paracrine fashion. Unexpectedly, IL-6 was also required for the execution of OIS, but in a cell-autonomous mode. Its depletion caused the inflammatory network to collapse and abolished senescence entry and maintenance. Furthermore, we demonstrate that the transcription factor C/EBPbeta cooperates with IL-6 to amplify the activation of the inflammatory network, including IL-8. In human colon adenomas, IL-8 specifically colocalized with arrested, p16(INK4A)-positive epithelium. We propose a model in which the context-dependent cytostatic and promitogenic functions of specific interleukins contribute to connect senescence with an inflammatory phenotype and cancer.

Simultaneous Isolation of DNA and RNA from the Same Cell Population Obtained by Laser Capture Microdissection for Genome and Transcriptome Profiling

Laser capture microdissection (LCM) is used extensively for genome and transcriptome profiling. Traditionally, however, DNA and RNA are purified from separate populations of LCM-harvested cells, limiting the strength of inferences about the relationship between gene expression and gene sequence variation. There have been no published protocols for the simultaneous isolation of DNA and RNA from the same cells that are obtained by LCM of patient tissue specimens. Here we report an adaptation of the Qiagen AllPrep method that allows the purification of DNA and RNA from the same LCM-harvested cells. We compared DNA and RNA purified by the QIAamp DNA Micro kit and the PicoPure RNA Isolation kit, respectively, from LCM-collected cells from adjacent tissue sections of the same specimen. The adapted method yields 90% of DNA and 38% of RNA compared with the individual methods. When tested with the GeneChip 250K Nsp Array, the concordance rate of the single nucleotide polymorphism heterozygosity calls was 98%. When tested with the GeneChip U133 Plus 2.0 Array, the correlation coefficient of the raw gene expression was 97%. Thus, we developed a method to obtain both DNA and RNA material from a single population of LCM-harvested cells and herein discuss the strengths and limitations of this methodology.

A Molecular Signature of Gastric Metaplasia Arising in Response to Acute Parietal Cell Loss

Loss of gastric parietal cells is a critical precursor to gastric metaplasia and neoplasia. However, the origin of metaplasia remains obscure. Acute parietal cell loss in gastrin-deficient mice treated with DMP-777 leads to the rapid emergence of spasmolytic polypeptide/trefoil factor family 2 (TFF2)-expressing metaplasia (SPEM) from the bases of fundic glands. We now sought to characterize more definitively the pathway for emergence of SPEM. Emerging SPEM lineages in gastrin-deficient mice treated with DMP-777 were examined for immunolocalization of TFF2, intrinsic factor, and Mist1, and morphologically with electron microscopy. Emerging SPEM was isolated with laser-capture microdissection and RNA was analyzed using gene microarrays. Immunohistochemistry in mouse and human samples was used to confirm up-regulated transcripts. DMP-777-induced SPEM was immunoreactive for TFF2 and the differentiated chief cell markers, Mist1 and intrinsic factor, suggesting that SPEM derived from transdifferentiation of chief cells. Microarray analysis of microdissected SPEM lineages induced by DMP-777 showed up-regulation of transcripts associated with G1/S cell-cycle transition including minichromosome maintenance deficient proteins, as well as a number of secreted factors, including human epididymis 4 (HE4). HE4, which was absent in the normal stomach, was expressed in SPEM of human and mouse and in intestinal metaplasia and gastric cancer in human beings. Although traditionally metaplasia was thought to originate from normal mucosal progenitor cells, these studies indicate that SPEM evolves through either transdifferentiation of chief cells or activation of a basal cryptic progenitor. In addition, induction of metaplasia elicits the expression of secreted factors, such as HE4, relevant to gastric preneoplasia.

Survival of Genetic Subtypes of Relapsed Myeloma May Be Modulated by Secondary Events.

The biological heterogeneity of myeloma can be largely explained by genetic subtypes characterized by primary translocations (t(4;14), t(11;14), and t(14;16)) or trisomies of odd-numbered chromosomes (hyperdiploid myeloma). In newly diagnosed patients, it is well established that these genetic subtypes have different survival. However, the prevalence and survival of the main genetic subtypes in relapse patients is unknown. In this study, we examined the prevalence and survival according to TC classes (a gene expression-based classification system which identifies the main genetic subtypes of myeloma) in patients with relapsed myeloma entered into bortezomib trials who had gene expression profiling (GEP) performed as part of a pharmacogenomics study (Bortezomib cohort). In comparison, we analyzed a large publicly available GEP dataset of newly diagnosed patients treated with total therapy II from the University of Arkansas for Medical Science (UAMS cohort, GEP data available on Gene Expression Omnibus, Zhan et al Blood 2006 epub). GEP was performed using the U133A and B chip for the bortezomib cohort and U133plus chip for the UAMS cohorts (Affymetrix, Santa Clara, CA). For both dataset, raw gene expression were log-transformed, median centered and analyzed using Genespring 7.0 (Agilent Technologies, Palo Alto, CA). A proliferation index (PI) based on the expression of 7 genes (TYMS, TOP2A, CCNB1, KIAA0101, CKS1B, UBE2C, TRIP13) was also calculated. As shown previously (Mulligan et al ASH 2005 abstract) the prevalence of the different TC classes in relapse patients was very similar to newly diagnosed patients (Chi-square p=0.25). In the bortezomib cohort, patients with t(11;14) had the worst survival when compared to patients with t(4;14), maf translocation and hyperdiploidy who had similar survival (survival of t(11;14) versus others, log-rank p=0.036). In contrast, among newly diagnosed patients (UAMS cohort) those with t(4;14) and maf translocations had the worst prognosis, whereas t(11;14) and hyperdiploid patients had the best prognosis (log-rank p=0.02). This interesting contrast in the prognosis within these main genetic subtypes of myeloma at diagnosis and relapse suggest that at relapse, the survival of these patients may be modulated by secondary events or possibly the treatment received. Indeed, a greater proportion of relapsed t(11;14) have a high PI (Chi-square p = 0.04) than newly diagnosed t(11;14). In contrast, there is no difference between the PI of newly diagnosed and relapsed hyperdiploid MM (TC D1, p=0.19). Our observations suggest that although the composition of the various genetic subtypes at diagnosis and relapse appear similar, the prognosis of some subtypes (especially t(11;14)) can be markedly different at relapse when modulated by secondary events (e.g. increase proliferation) whereas others (such as hyperdiploid MM) appear to behave similarly at diagnosis and relapse. Our observation also has important therapeutic implications as it suggests relapsed t(11;14) should be treated aggressively and drug combinations or addition of new agents may be more appropriate. Conversely, bortezomib seemed to be an effective therapy in relapse t(4;14) patients, improving the prognosis of these patients to that of genetic subgroups with good prognosis. Our study highlights the fact that different treatment strategies may be needed for newly diagnosed and relapsed patients even within same genetic subtypes.