Search

Abstract Background Network inference from high-throughput data has become an important means of current analysis of biological systems. For instance, in cancer research, the functional relationships of cancer related proteins, summarised into signalling networks are of central interest for the identification of pathways that influence tumour development. Cancer cell lines can be used as model systems to study the cellular response to drug treatments in a time-resolved way. Based on these kind of data, modelling approaches for the signalling relationships are needed, that allow to generate hypotheses on potential interference points in the networks. Results We present the R-package 'ddepn' that implements our recent approach on network reconstruction from longitudinal data generated after external perturbation of network components. We extend our approach by two novel methods: a Markov Chain Monte Carlo method for sampling network structures with two edge types (activation and inhibition) and an extension of a prior model that penalises deviances from a given reference network while incorporating these two types of edges. Further, as alternative prior we include a model that learns signalling networks with the scale-free property. Conclusions The package 'ddepn' is freely available on R-Forge and CRAN http://ddepn.r-forge.r-project.org, http://cran.r-project.org. It allows to conveniently perform network inference from longitudinal high-throughput data using two different sampling based network structure search algorithms.

Currently there are no sufficiently sensitive biomarkers able to reflect changes in joint remodelling during osteoarthritis (OA). In this work, we took an affinity proteomic approach to profile serum samples for proteins that could serve as indicators for the diagnosis of radiographic knee OA. Antibody suspension bead arrays were applied to analyze serum samples from patients with OA (n = 273), control subjects (n = 76) and patients with rheumatoid arthritis (RA, n = 244). For verification, a focused bead array was built and applied to an independent set of serum samples from patients with OA (n = 188), control individuals (n = 83) and RA (n = 168) patients. A linear regression analysis adjusting for sex, age and body mass index (BMI) revealed that three proteins were significantly elevated (P < 0.05) in serum from OA patients compared to controls: C3, ITIH1 and S100A6. A panel consisting of these three proteins had an area under the curve of 0.82 for the classification of OA and control samples. Moreover, C3 and ITIH1 levels were also found to be significantly elevated (P < 0.05) in OA patients compared to RA patients. Upon validation in additional study sets, the alterations of these three candidate serum biomarker proteins could support the diagnosis of radiographic knee OA.

There is a demand for novel targets and approaches to diagnose and treat prostate cancer (PCA). In this context, serum and plasma samples from a total of 609 individuals from two independent patient cohorts were screened for IgG reactivity against a sum of 3833 human protein fragments. Starting from planar protein arrays with 3786 protein fragments to screen 80 patients with and without PCA diagnosis, 161 fragments (4%) were chosen for further analysis based on their reactivity profiles. Adding 71 antigens from literature, the selection of antigens was corroborated for their reactivity in a set of 550 samples using suspension bead arrays. The antigens prostein (SLC45A3), TATA-box binding protein (TBP), and insulin-like growth factor 2 mRNA binding protein 2 (IGF2BP2) showed higher reactivity in PCA patients with late disease compared with early disease. Because of its prostate tissue specificity, we focused on prostein and continued with mapping epitopes of the 66-mer protein fragment using patient samples. Using bead-based assays and 15-mer peptides, a minimal peptide epitope was identified and refined by alanine scanning to the KPxAPFP. Further sequence alignment of this motif revealed homology to transmembrane protein 79 (TMEM79) and TGF-beta-induced factor 2 (TGIF2), thus providing a reasoning for cross-reactivity found in females. A comprehensive workflow to discover and validate IgG reactivity against prostein and homologous targets in human serum and plasma was applied. This study provides useful information when searching for novel biomarkers or drug targets that are guided by the reactivity of the immune system against autoantigens.

Background: Cancer cachexia (CC) is linked to poor prognosis. Although the mechanisms promoting this condition are not known, several circulating proteins have been proposed to contribute. We analyzed the plasma proteome in cancer subjects in order to identify factors associated with cachexia., Design/subjects: Plasma was obtained from a screening cohort of 59 patients, newly diagnosed with suspected gastrointestinal cancer, with (n = 32) or without (n = 27) cachexia. Samples were subjected to proteomic profiling using 760 antibodies (targeting 698 individual proteins) from the Human Protein Atlas project. The main findings were validated in a cohort of 93 patients with verified and advanced pancreas cancer., Results: Only six proteins displayed differential plasma levels in the screening cohort. Among these, Carnosine Dipeptidase 1 (CNDP1) was confirmed by sandwich immunoassay to be lower in CC (p = 0.008). In both cohorts, low CNDP1 levels were associated with markers of poor prognosis including weight loss, malnutrition, lipid breakdown, low circulating albumin/IGF1 levels and poor quality of life. Eleven of the subjects in the discovery cohort were finally diagnosed with non-malignant disease but omitting these subjects from the analyses did not have any major influence on the results., Conclusions: In gastrointestinal cancer, reduced plasma levels of CNDP1 associate with signs of catabolism and poor outcome. These results, together with recently published data demonstrating lower circulating CNDP1 in subjects with glioblastoma and metastatic prostate cancer, suggest that CNDP1 may constitute a marker of aggressive cancer and CC.

Osteoarthritis (OA) is the most common rheumatic disease and one of the most disabling pathologies worldwide. To date, the diagnostic methods of OA are very limited, and there are no available medications capable of halting its characteristic cartilage degeneration. Therefore, there is a significant interest in new biomarkers useful for the early diagnosis, prognosis, and therapeutic monitoring. In the recent years, protein microarrays have emerged as a powerful proteomic tool to search for new biomarkers. In this study, we have used two concepts for generating protein arrays, antigen microarrays, and NAPPA (nucleic acid programmable protein arrays), to characterize differential autoantibody profiles in a set of 62 samples from OA, rheumatoid arthritis (RA), and healthy controls. An untargeted screen was performed on 3840 protein fragments spotted on planar antigen arrays, and 373 antigens were selected for validation on bead-based arrays. In the NAPPA approach, a targeted screening was performed on 80 preselected proteins. The autoantibody targeting CHST14 was validated by ELISA in the same set of patients. Altogether, nine and seven disease related autoantibody target candidates were identified, and this work demonstrates a combination of these two array concepts for biomarker discovery and their usefulness for characterizing disease-specific autoantibody profiles.

Background: Despite promising progress in targeted breast cancer therapy, drug resistance remains challenging. The monoclonal antibody drugs trastuzumab and pertuzumab as well as the small molecule inhibitor erlotinib were designed to prevent ErbB-2 and ErbB-1 receptor induced deregulated protein signalling, contributing to tumour progression. The oncogenic potential of ErbB receptors unfolds in case of overexpression or mutations. Dimerisation with other receptors allows to bypass pathway blockades. Our intention is to reconstruct the ErbB network to reveal resistance mechanisms. We used longitudinal proteomic data of ErbB receptors and downstream targets in the ErbB-2 amplified breast cancer cell lines BT474, SKBR3 and HCC1954 treated with erlotinib, trastuzumab or pertuzumab, alone or combined, up to 60 minutes and 30 hours, respectively. In a Boolean modelling approach, signalling networks were reconstructed based on these data in a cell line and time course specific manner, including prior literature knowledge. Finally, we simulated network response to inhibitor combinations to detect signalling nodes reflecting growth inhibition., Results: The networks pointed to cell line specific activation patterns of the MAPK and PI3K pathway. In BT474, the PI3K signal route was favoured, while in SKBR3, novel edges highlighted MAPK signalling. In HCC1954, the inferred edges stimulated both pathways. For example, we uncovered feedback loops amplifying PI3K signalling, in line with the known trastuzumab resistance of this cell line. In the perturbation simulations on the short-term networks, we analysed ERK1/2, AKT and p70S6K. The results indicated a pathway specific drug response, driven by the type of growth factor stimulus. HCC1954 revealed an edgetic type of PIK3CA-mutation, contributing to trastuzumab inefficacy. Drug impact on the AKT and ERK1/2 signalling axes is mirrored by effects on RB and RPS6, relating to phenotypic events like cell growth or proliferation. Therefore, we additionally analysed RB and RPS6 in the long-term networks., Conclusions: We derived protein interaction models for three breast cancer cell lines. Changes compared to the common reference network hint towards individual characteristics and potential drug resistance mechanisms. Simulation of perturbations were consistent with the experimental data, confirming our combined reverse and forward engineering approach as valuable for drug discovery and personalised medicine.

Non-alcoholic fatty liver disease comprises a broad spectrum of disease states ranging from simple steatosis to non-alcoholic steatohepatitis. As a result of increases in the prevalences of obesity, insulin resistance, and hyperlipidemia, the number of people with hepatic steatosis continues to increase. Differences in susceptibility to steatohepatitis and its progression to cirrhosis have been attributed to a complex interplay of genetic and external factors all addressing the intracellular network. Increase in sugar or refined carbohydrate consumption results in an increase of insulin and insulin resistance that can lead to the accumulation of fat in the liver. Here we demonstrate how a multidisciplinary approach encompassing cellular reprogramming, transcriptomics, proteomics, metabolomics, modeling, network reconstruction, and data management can be employed to unveil the mechanisms underlying the progression of steatosis. Proteomics revealed reduced AKT/mTOR signaling in fibroblasts derived from steatosis patients and further establishes that the insulin-resistant phenotype is present not only in insulin-metabolizing central organs, e.g., the liver, but is also manifested in skin fibroblasts. Transcriptome data enabled the generation of a regulatory network based on the transcription factor SREBF1, linked to a metabolic network of glycerolipid, and fatty acid biosynthesis including the downstream transcriptional targets of SREBF1 which include LIPIN1 (LPIN) and low density lipoprotein receptor. Glutathione metabolism was among the pathways enriched in steatosis patients in comparison to healthy controls. By using a model of the glutathione pathway we predict a significant increase in the flux through glutathione synthesis as both gamma-glutamylcysteine synthetase and glutathione synthetase have an increased flux. We anticipate that a larger cohort of patients and matched controls will confirm our preliminary findings presented here.

Increasing the efficacy of targeted cancer therapies requires the identification of robust biomarkers suitable for patient stratification. This study focused on the identification of molecular mechanisms causing resistance against the anti-ERBB2-directed therapeutic antibodies trastuzumab and pertuzumab presently used to treat patients with ERBB2-amplified breast cancer. Immunohistochemistry and clinical data were evaluated and yielded evidence for the existence of ERBB2-amplified breast cancer with high-level epidermal growth-factor receptor (EGFR) expression as a separate tumor entity. Because the proto-oncogene EGFR tightly interacts with ERBB2 on the protein level, the hypothesis that high-level EGFR expression might contribute to resistance against ERBB2-directed therapies was experimentally validated. SKBR3 and HCC1954 cells were chosen as model systems of EGFR-high/ERBB2-amplified breast cancer and exposed to trastuzumab, pertuzumab and erlotinib, respectively, and in combination. Drug impact was quantified in cell viability assays and on the proteomic level using reverse-phase protein arrays. Phosphoprotein dynamics revealed a significant downregulation of AKT signaling after exposure to trastuzumab, pertuzumab or a coapplication of both antibodies in SKBR3 cells but no concomitant impact on ERK1/2, RB or RPS6 phosphorylation. On the other hand, signaling was fully downregulated in SKBR3 cells after coinhibition of EGFR and ERBB2. Inhibitory effects in HCC1954 cells were driven by erlotinib alone, and a significant upregulation of RPS6 and RB phosphorylation was observed after coincubation with pertuzumab and trastuzumab. In summary, proteomic data suggest that high-level expression of EGFR in ERBB2-amplified breast cancer cells attenuates the effect of anti-ERBB2-directed antibodies. In conclusion, EGFR expression may serve as diagnostic and predictive biomarker to advance personalized treatment concepts of patients with ERBB2-amplified breast cancer.

The EGFR-driven cell-cycle pathway has been extensively studied due to its pivotal role in breast cancer proliferation and pathogenesis. Although several studies reported regulation of individual pathway components by microRNAs (miRNAs), little is known about how miRNAs coordinate the EGFR protein network on a global miRNA (miRNome) level. Here, we combined a large-scale miRNA screening approach with a high-throughput proteomic readout and network-based data analysis to identify which miRNAs are involved, and to uncover potential regulatory patterns. Our results indicated that the regulation of proteins by miRNAs is dominated by the nucleotide matching mechanism between seed sequences of the miRNAs and 3'-UTR of target genes. Furthermore, the novel network-analysis methodology we developed implied the existence of consistent intrinsic regulatory patterns where miRNAs simultaneously co-regulate several proteins acting in the same functional module. Finally, our approach led us to identify and validate three miRNAs (miR-124, miR-147 and miR-193a-3p) as novel tumor suppressors that co-target EGFR-driven cell-cycle network proteins and inhibit cell-cycle progression and proliferation in breast cancer.

Background: Network inference from high-throughput data has become an important means of current analysis of biological systems. For instance, in cancer research, the functional relationships of cancer related proteins, summarised into signalling networks are of central interest for the identification of pathways that influence tumour development. Cancer cell lines can be used as model systems to study the cellular response to drug treatments in a time-resolved way. Based on these kind of data, modelling approaches for the signalling relationships are needed, that allow to generate hypotheses on potential interference points in the networks., Results: We present the R-package 'ddepn' that implements our recent approach on network reconstruction from longitudinal data generated after external perturbation of network components. We extend our approach by two novel methods: a Markov Chain Monte Carlo method for sampling network structures with two edge types (activation and inhibition) and an extension of a prior model that penalises deviances from a given reference network while incorporating these two types of edges. Further, as alternative prior we include a model that learns signalling networks with the scale-free property., Conclusions: The package 'ddepn' is freely available on R-Forge and CRAN http://ddepn.r-forge.r-project.org, http://cran.r-project.org. It allows to conveniently perform network inference from longitudinal high-throughput data using two different sampling based network structure search algorithms.

Protein microarrays are an ideal technology platform which allow for a robust and standardized profiling of the cellular proteome. Many cellular functions are not simply controlled by the presence of certain proteins, especially the propagation of external stimuli, which depend on transient post-translational modifications that determine whether a protein is in its active or inactive state. Thus, complex biological processes require the availability of a sound set of quantitative and time-resolved measurements to be understood. For this reason, new assay platforms which allow for the investigation of several proteins in parallel are necessary. The current best understood mode of cellular regulation occurs via phosphorylation and dephosphorylation processes, which are mediated via a large panel of kinases and phosphatases. The microspot immunoassay technique described here allows for an exact determination of several different phosphorylated proteins in parallel, as well as from small sample amounts, and is therefore an appropriate system to deepen the understanding of the complex regulatory networks implicated in health and disease.

Unlabelled: Protein microarrays are well-established as sensitive tools for proteomics. Particularly, the microspot immunoassay (MIA) platform enables a quantitative analysis of (phospho-) proteins in complex solutions (e.g. cell lysates or blood plasma) and with low consumption of samples and reagents. Despite numerous biological and clinical applications of MIAs there is currently no user-friendly open source data analysis software available with versatile options for data analysis and data visualization. Here, we introduce the open source software QuantProReloaded that is specifically designed for the analysis of data from MIA experiments., Availability and Implementation: QuantProReloaded is written in R and Java and is open for download under the BSB license at http://code.google.com/p/quantproreloaded/.

Motivation: Network modelling in systems biology has become an important tool to study molecular interactions in cancer research, because understanding the interplay of proteins is necessary for developing novel drugs and therapies. De novo reconstruction of signalling pathways from data allows to unravel interactions between proteins and make qualitative statements on possible aberrations of the cellular regulatory program. We present a new method for reconstructing signalling networks from time course experiments after external perturbation and show an application of the method to data measuring abundance of phosphorylated proteins in a human breast cancer cell line, generated on reverse phase protein arrays., Results: Signalling dynamics is modelled using active and passive states for each protein at each timepoint. A fixed signal propagation scheme generates a set of possible state transitions on a discrete timescale for a given network hypothesis, reducing the number of theoretically reachable states. A likelihood score is proposed, describing the probability of measurements given the states of the proteins over time. The optimal sequence of state transitions is found via a hidden Markov model and network structure search is performed using a genetic algorithm that optimizes the overall likelihood of a population of candidate networks. Our method shows increased performance compared with two different dynamical Bayesian network approaches. For our real data, we were able to find several known signalling cascades from the ERBB signalling pathway., Availability: Dynamic deterministic effects propagation networks is implemented in the R programming language and available at http://www.dkfz.de/mga2/ddepn/.

Summary: RPPanalyzer is a statistical tool developed to read reverse-phase protein array data, to perform the basic data analysis and to visualize the resulting biological information. The R-package provides different functions to compare protein expression levels of different samples and to normalize the data. Implemented plotting functions permit a quality control by monitoring data distribution and signal validity. Finally, the data can be visualized in heatmaps, boxplots, time course plots and correlation plots. RPPanalyzer is a flexible tool and tolerates a huge variety of different experimental designs., Availability: The RPPAanalyzer is open source and freely available as an R-Package on the CRAN platform http://cran.r-project.org/.

The quantitative analysis of signaling networks requires highly sensitive methods for the time-resolved determination of protein phosphorylation. For this reason, we developed a quantitative protein microarray that monitors the activation of multiple signaling pathways in parallel, and at high temporal resolution. A label-free sandwich approach was combined with near infrared detection, thus permitting the accurate quantification of low-level phosphoproteins in limited biological samples corresponding to less than 50,000 cells, and with a very low standard deviation of approximately 5%. The identification of suitable antibody pairs was facilitated by determining their accuracy and dynamic range using our customized software package Quantpro. Thus, we are providing an important tool to generate quantitative data for systems biology approaches, and to drive innovative diagnostic applications.

A significant bottleneck for the time-resolved and quantitative description of signaling networks is the limited sample capacity and sensitivity of existing methods. Recently, antibody microarrays have emerged as a promising experimental platform for the quantitative and comprehensive determination of protein abundance and protein phosphorylation. This review summarizes the development of microarray applications involving antibody-based capture of target proteins with a focus on quantitative applications. Technical aspects regarding the production of antibody microarrays, identification of suitable detection and capture antibody pairs, signal detection methods, detection limit, and data analysis are discussed in detail.

The advancement of efficient technologies to comply with the needs of systems biology and drug discovery has so far not received adequate attention. A substantial bottleneck for the time-resolved quantitative description of signaling networks is the limited throughput and the inadequate sensitivity of currently established methods. Here, we present an improved protein microarray-based approach towards the sensitive detection of proteins in the fg-range which is based on signal detection in the near-infrared range. The high sensitivity of the assay permits the specific quantification of proteins derived from as little as only 20,000 cells with an error rate of only 5%. The capacity is limited to the analysis of up to 500 different samples per microarray. Protein abundance is determined qualitatively, and quantitatively, if recombinant protein is available. This novel approach was called IPAQ (infrared-based protein arrays with quantitative readout). IPAQ offers a highly sensitive experimental approach superior to the established standard protein quantification technologies, and is suitable for quantitative proteomics. Employing the IPAQ approach, a detailed analysis of activated signaling networks in biopsy samples and of crosstalk between signaling modules as required in drug discovery strategies can easily be performed.