Final Report Summary - SYSVASC (Systems Biology to Identify Molecular Targets for Vascular Disease Treatment)
Executive Summary:
1.1. Executive Summary
“Systems Biology to Identify Molecular Targets for Vascular Disease Treatment” (sysVASC) was a collaborative project based on established multidisciplinary European research networks, including specialists in pre-clinical and clinical research, omics technologies, and systems biology from research intensive SMEs and academia. sysVASC was funded through the European Union’s Seventh Framework Programme (FP7- Collaborative Project (CP) - Small or medium-scale focused research project (STREP)) and was officially launched in February 2014 and ran for 48 months. The project website can be found at http://www.sysvasc.eu/.
Cardiovascular disease (CVD) is the leading cause of mortality and morbidity in Europe and worldwide. Asymptomatic vascular damage accumulates for years before patients are identified and subjected to therapeutic measures. The limited knowledge on early vascular disease pathophysiology is reflected in the lack of therapeutic options. Prognostic biomarkers still remain widely elusive and personalised risk assessment relies on rough estimates based on traditional risk factors. The underlying pathophysiological changes remain incompletely understood. The main clinical needs faced in CVD are: (i) a better understanding of the key pathogenic factors involved in the progression of CVD independent of aetiology, (ii) the diagnosis of asymptomatic disease and (iii) the prediction and prevention of cardiovascular events.
sysVASC aimed to overcome this limitation by identifying pathophysiological mechanisms and key molecules responsible for onset and progression of CVD and validation of their potential to serve as molecular targets for therapeutic intervention. The sysVASC consortium gathering a wide range of experts in the field of different “omics” technologies, systems biology, statistics, cardiovascular diseases and preclinical studies provided a unique know-how necessary to break through current challenges in the field of CVD. As a result, the molecular phenotype of progressive CVD was identified using high resolution “omics” data in both humans and animal models. We collected multilevel data of 1949 individuals with CVD that in a systems biology analysis integrating clinical features, tissue and body fluid multi-omics data and extensive literature data led to the identification of a number of original targets in CVD focussing on calcification, extracellular matrix and epigenetic modifications. In parallel a wide range of animal CVD models were characterised to the same molecular level as to determine which of the animal models represent optimally human CVD. These models were used in experiments aiming to interfere with sysVASC targets leading to confirmation of the relevance of these targets in CVD and their suitability for further drug-development.
Project Context and Objectives:
1.2. Project Context and Objectives
1.2.1 Project context
Cardiovascular disease (CVD) is the leading cause of mortality and morbidity in Europe and worldwide. The term CVD comprises a multitude of clinical conditions and different degrees of severity. The pathophysiology of CVD is complex and multifactorial and the limited knowledge on early vascular disease pathophysiology is reflected in the lack of therapeutic options. Before patients are diagnosed and treated, asymptomatic vascular damage can accumulate for years. Although early pathophysiological changes are at least in part reversible, their diagnosis remains challenging. They are clinically asymptomatic and require specialised equipment to characterise the vascular phenotype, and a universally accepted method to assess vascular function and structure in the clinical context is not available. The main clinical needs faced in CVD are: a) a better understanding of the key pathogenic factors involved in the progression of CVD independent of aetiology, b) the diagnosis of asymptomatic disease and c) the prediction and prevention of cardiovascular events. Prognostic molecular biomarkers remain widely elusive and clinical practice currently relies on rough estimates based on traditional risk factors. Consequently, current preventative measures are of general nature and aim at these risk factors without taking their relative contribution to an individual's pathophysiology into account. It is presumed that a clearer understanding of the pathophysiology of CVD will promote personalised risk assessment and targeted preventative and therapeutic measures.
The sysVASC Consortium aimed to overcome this limitation by mounting a comprehensive systems medicine approach to elucidate pathological mechanisms, which will yield molecular targets for therapeutic intervention. The sysVASC Consortium compiled the multi-disciplinary skills and unique expertise necessary to tackle this ambitious task. We integrated high quality, high resolution molecular omics data from clinical studies in order to comprehensively describe the molecular phenotype of progressive CVD. Statistical modelling identified targets for specific interventions. These have been initially validated in models best mimicking key pathogenic processes in human disease.
1.2.2 Project Objectives
The main objective of the sysVASC project was to identify key factors involved in the pathogenesis and progression of coronary artery disease (CAD) and to identify molecular targets for the treatment of CVD. The project had 5 more specific objectives that have been addressed throughout the lifetime of sysVASC.
Objective 1: To identify the key pathogenic factors involved in the progression of CAD and leading to coronary events.
The sysVASC consortium decided to focus its research on macrovascular disease. For literature-based work CAD has been chosen as model disease due to the very clear definition of the two key clinical events: myocardial infarction and death from a cardiovascular event. In experiments in experimental models and human tissue samples the consortium benefited from detailed histological analyses and/or functional data on vascular phenotypes. In large cohort data we applied a specifically developed staging model in order to estimate severity of vascular disease.
Objective 2: To study the molecular features of CAD independently of aetiology.
Pathogenic factors including hypertension, diabetes, smoking and dyslipidaemia trigger different processes in the initiation of vascular diseases. In more advanced, yet still clinically asymptomatic stages, the progression is in part independent of the original aetiology. The sysVASC consortium collected therefore well phenotyped clinical samples that were used for in depth multi-omics molecular analysis (transcriptomics, proteomics/peptidomics, targeted and untargeted metabolomics) and for the integration of clinical, pathological and molecular phenotypes towards understanding the CVD pathophysiology.
Objective 3: To apply systems medicine approaches to define the key nodes of pathogenic networks involved in the progression of CVD.
Complex multifactorial and chronic pathologies such as CVD cannot be described on a single molecule level. Instead, a comprehensive analysis needs to uncover nodes within a complex framework of interactions, which maintain disease mechanisms. sysVASC thus developed an ambitious systems medicine approach to uncover molecular pathways involved in the progression of macrovascular disease from early stages to clinical events. These key factors and pathways identified in human studies were then targeted in the most appropriate experimental models to prove their role in disease progression and as targets for preventative and therapeutic approaches.
Objective 4: To identify animal models mimicking specific key elements of human CVD.
In parallel to human data, the sysVASC consortium carried out an in-depth characterisation of CVD development in animal models representing the key risk factors, including hypertension, diabetes, hypercholesterolaemia, hypercalcemia and combinations thereof. This suite of animal CVD models were characterised to the same level of detail as human samples by applying proteomics, peptidomics and metabolomics to the animal body fluids and tissue, enabling precise identification of common features between animal models and humans. This led to the definition of the most appropriate models representing certain aspects of human CVD for subsequent molecular and interference studies. A number of these animal models combined with the massive omics data also served as starting points of novel CVD targets.
Objective 5: To validate new therapeutic targets hampering the progression of CVD in the most appropriate animal models.
The systems biology analysis components of sysVASC lead to the identification of a number of novel CVD targets and their associated compounds for interference with their activity. Other computational approaches based on the sysVASC omics data lead to repurposing of compounds. Using animal models best representing human CVD, we carried in vivo interference studies using these compounds to determine their suitability as candidates for drug development. For a number of compounds we obtained in vitro data supporting their potential as targets in CVD. Very diverse and novel CVD targets were validated with a specific focus on calcification, ECM and epigenetic modifications
Project Results:
1.3. The sysVASC Work Packages and Major Scientific Results
sysVASC was organised in five research based Work Packages (WP1-5) and three Work Packages on exploitation, dissemination and project management (WP6-8) (Figure 1).
Figure 1: sysVASC Work Packages and their interaction.
WP1 focussed on (i) the establishment and maintenance of a central data repository to deposit and integrate all available phenotypic and omics data; (ii) the development of a new CVD knowledge base containing information extracted from existing databases and literature searches; (iii) systemic analysis of (a) existing data and (b) new data on human samples generated within the consortium.
The main task of WP2 was (i) the generation of human CVD omics datasets focussing on human vascular tissue and (ii) the verification of changes of novel targets predicted by the in silico models.
WP3 used the existing and newly generated CVD data for (i) statistical modelling; (ii) identification of association networks and key nodes; (iii) advanced system biology modelling including inter-species comparison.
Within WP4 existing animal models of the key risk factors for CVD (hypertension, hypercholesterolemia, diabetes, and hypercalcemia) have been (i) molecularly characterised using omics technologies; (ii) evaluated regarding their similarity to the human disease.
The nature of the compounds to be tested within WP5 was in the beginning unknown, as this depended on the data from WP3/WP4. WP5 (i) tested the novel targets identified within the other WPs in interference studies in CVD models; and (ii) defined the molecular mechanism of action of those interfering compounds.
Key achievements of the Work Packages are highlighted in the following section.
1.3.1. WP1: Existing multi-source data integration from human disease
The sysVASC consortium benefited from human datasets that were brought into the project by consortium partners. A number of large-scale cohort data such as the “Kooperative Gesundheitsforschung in der Region Augsburg" (Cooperative Health Research in the Augsburg Region; KORA) cohort with associated omics datasets were available for in-depth analysis in WP1 and generation of additional data in other work packages. Not all of these clinical datasets have been generated primarily to study vascular diseases. Moreover, specific vascular phenotypes such as data on endothelial function, vascular stiffness, atherosclerosis; imaging data including coronary angiography; and tissue samples for detailed histological analysis of vascular damage were not consistently available in these cohorts. This is a common feature of such large datasets and the consortium was well aware of the challenges related to discrepancies between the availability of precise molecular phenotypes and less detailed and particularly, unstandardized phenotypic characterisation across the clinical datasets. These challenges extend further to literature data that may contain relevant information but cannot immediately be integrated with other data due to lack of standardisation of phenotypes.
WP1 therefore employed a number of strategies to facilitate data integration across clinical datasets that were directly available to the consortium and datasets that were obtained from literature and publicly available database searches. The key achievements of WP1 in this respect are:
• Development of a classification system of severity of vascular disease based on commonly available risk factors and clinical features.
• Development of a consortium-wide data repository to facilitate integration of data across human studies and also across other work packages. From here a more sophisticated platform has been developed to support the wider research community in their data integration efforts.
• Development of an integrated database featuring clinical and molecular data across human cohorts.
• Development of an online database of molecular features associated with vascular disease based on available literature data.
The classification system for vascular diseases has been developed to grade vascular disease severity in cohorts that do not have specific vascular phenotypic information available and to allow comparison between cohorts that employed different approaches to characterise vascular disease. It is based on other grading systems that are in clinical use, e.g. for the severity of other cardiovascular disease (Figure 2).
Figure 2: Classification system for severity of vascular diseases.
The system is based on 5 stages from A (not at risk for and no evidence of established vascular disease) to E (evidence of structural vascular disease). For practical reasons it is a combination of risk assessment and assessment of established disease and therefore remains an element of uncertainty. The grading system has therefore been used throughout the duration of the project for analysis of large cohorts where inherent imprecision can be compensated by large numbers and for screening experiments where a first selection of samples was based on the grading, followed by in-depth analysis of specific vascular phenotypes and/or histology. Where required the higher grades (D/E) were combined and contrasted against lower grades (A/B).
In parallel to this clinical grading system the consortium has made use of specific vascular phenotypes including quantitative coronary angiography, pulse wave velocity and carotid intima-media thickness. A parallel histological grading system has also been developed and applied where tissue samples were available: A: completely thin intima; B: intima slightly thickened; C: intima mostly thickened and partly strong thickened; D: thickened intima + calcification in intima; E: thickened intima + calcification in media. This system played a more crucial role for other work packages, in particular for WP2 (Figure 3).
Figure 3: Samples of common iliac arteries from a patient with histological grade A (left panel) and a patient with histological grade E (right panel) vascular disease.
In order to make all project data accessible to project participants a dedicated data repository has been developed. Secure data sharing between consortium members has been facilitated by means of an sFTP (FTP over ssh) managed server. Information security compliance is achieved by having SSL certificate encryption, regular backups and 24/7 monitoring by the server provider. Storage capability is suitable to the expected size of the project. The server has been administered by Eagle Genomics and users accessed the site via password control that has been communicated to a restricted number of consortium partners. The sFTP site has been available for use by all sysVASC partners and access to the data will be maintained beyond the funding period.
Based on the experience within sysVASC Eagle Genomics have been developing the e[automateddatascientst] (e[ADS]) platform. This platform aims to facilitate data integration and exploration in life sciences. It is a next generation conversational learning platform for data integration and discovery. The platform has 5 main modules that adhere to concepts concerning data cataloguing, data curation, data analysis, data valuation and genome visualisation. It is populated with data imported from various data sources that cover many research areas: genome, genes and transcripts from Ensembl; proteins from UniProt; pathways from Reactome; SNPs from dbSNP; association with disease from OpenTargets; pharmacology from DrugBank; and scientific literature from PubMed and EuroPMC. Data extracted from these public resources allow navigation between entities of interest, showing the evidence based relationships between each entity. This platform is constantly being developed in response to customer requirements, identifying new data sources and extending the data model with new entities. While the indication of interest for sysVASC is cardiovascular disease, the data model can be adapted for any disease area and is therefore attractive to many potential customers by delivering a range of business benefits such as improved productivity and collaboration which combined speed up time to insight.
In order to directly explore the available omics datasets across a number of clinical cohorts and within cohorts the sysVASC consortium required an additional solution that brings data together in an easily accessible format. Led by project partner Mosaiques Diagnostics the consortium has developed a database integrating resources from all partners within sysVASC that can be accessed online (http://sysvasc.mosaiques.de). Table 1 provides an overview of the large amount of data that were added to the database through regular updates.
Whilst these efforts focussed on data that are immediately available to the consortium there have been further efforts to bring literature data and data from public repositories into the sysVASC consortium. A new Cardiovascular Disease knowledge base (sysVASC knowledge base; http://cvdkb.org/) containing information extracted from existing databases and literature searches has therefore been developed based on an existing integrated knowledge base for kidney diseases (www.kupkb.org). The construction of the CVDKB involved the conversion of various external databases and required collection of experimental findings from high-throughput experiments investigating gene, protein and metabolite in the context of coronary artery disease. A decision has been made to focus on stable coronary artery disease in order to have a precise definition of a vascular disease phenotype available for implementation of the knowledge base. Additional features can be added in the future.
Table 1: Clinical and omics features contained in the sysVASC integrated database.
Features in the database Type of data Update 1 Update 2 Update 3 Update 4 Update 5 Update 6
DB_Patient Clinical 616 1,658 2,274 2,274 2,274 2,320
DB_Sample Clinical 1,074 2,111 3,184 3,184 3,184 3,184
DB_tbEvaluation Clinical 1,766 1,890 2,506 2,506 2,506 2,506
DB_OrigSample Clinical 631 1,659 2,289 2,289 2,289 2,477
DB_Analyses Clinical 1,028 2,239 3,267 3,267 3,267 3,267
DB_Diagnoses Clinical 845 1,746 2,591 2,591 2,591 3,739
DB_Parameters (biochemical) Clinical 3,579 15,366 18,945 18,945 18,945 21,872
DB_Drugs Clinical 0 0 0 0 1,502 1502
DB_CE_MS “Omics” 3,460,004 3,460,004 3,460,004 3,460,004 14,074,069 14,074,069
DB_LC_MS “Omics” 69,890 69,890 69,890 69,890 69,890 298,085
DB_Analyte_List “Omics” 0 0 0 0 0 451
DB_Analyte_Data “Omics” 0 0 0 0 0 525
The datasets present in the CVDKB can be divided into three groups (Figure 4): the domain ontologies, which add semantic annotation on top, to the experimental data to be queried; the omics database that helps linking the different biological layers; and the experimental data associated with coronary heart disease.
Figure 4: Schematic structure of CVDKB showing experimental data connected to background knowledge and annotated with the domain ontologies. Ontologies capture the relationships between entities and express these relationships in a language that can be interpreted by a computer. Ontologies are needed in order to provide a common vocabulary for describing cardiovascular anatomy, cells and diseases.
1.3.2. WP2: Generation of additional data and verification of predicted targets in human samples
The main objectives of WP2 included:
1) Collection of well phenotyped clinical samples in accordance to latest ethics regulations that can serve as an invaluable resource for in depth multi-omics molecular analysis and integration of clinical, pathological and molecular phenotypes towards understanding CVD pathophysiology.
2) Generation of multi-level omics datasets on human CVD to complement existing knowledge. This analysis included studies at all genome, transcriptome, proteome/peptidome and metabolome levels on tissue, plasma and urine, as applicable.
3) Verify differential expression of selected molecular features via a combination of tools including cross-omics comparisons, extensive literature mining as well as immunoassays.
1. Collection of clinical samples
By the end of sysVASC, 138 patients were included in the sysVASC study, with the number of vascular tissue samples (i.e. different vessels) per patient varying. Clinical data from patients are organised in a database structure to support statistical computation. The entire sample cohort is stored at CHA.
The Table below (Table 2) provides an overview of the samples contained in the collection until M48. The stages (A-E) are in accordance with the classification system established in WP1.
Table 2: Human samples available from de novo collection at MUG by the end of Period 3
Collection of omics datasets
Multiple new omics datasets were collected in the course of sysVASC. These included:
• Whole-exome-sequencing and expression profiling by RNA sequencing from a total of 33 tissue (n = 32 patients) samples.
• High resolution proteomics data from 43 plasma, 108 serum, and 70 vessel (n = 59 patients) samples.
• Urine peptidomics datasets from approximately 1000 samples (samples from the French and euRopean Outcome reGistry in Intensive Care Units (FROG-ICU) study were also included in the analysis).
• Targeted metabolomics datasets from 126 plasma, 108 urine and 66 tissue (n = 55 patients) samples. Untargeted metabolomics using FT-ICR from 149 tissue (n = 42 patients) samples, 43 matched blood plasma and 36 urine samples from the sysVASC cohort and 1000 blood and urine samples from the KORA cohort.
These data were analysed using a combination of tools including, cross-omics comparisons, literature mining, pathway analysis, and investigation of relevant existing data-sources, highlighting a series of interesting targets (described in WP4-WP5).
As an example of the comprehensive analysis conducted at the protein level, a snapshot from the comparisons of the results from central versus peripheral vessel analysis is shown below (Table 3).
Table 3: Examples of consistent proteomic differences between CVD cases and controls in central vessels, peripheral vessels and pooled (central plus peripheral vessel) analyses. Based on an extensive literature search all of the shown proteins have already been described in the context of CVD, supporting the validity of the approach. Red colour indicates up-; green colour: down-regulation in cases versus controls. In blue: the difference reached statistical significance in the particular comparison.
Symbol Name central vessels peripheral vessels pooled analysis
Number of peptides ratio Mann Whitney p-value Number of peptides ratio Mann Whitney p-value Number of peptides ratio Mann Whitney p-value
PDLIM7 PDZ and LIM domain protein 7 3 0.42 3.70E-03 3 0.67 5.55E-02 2 0.56 2.33E-03
HSPB6 Heat shock protein beta-6 2 0.77 6.87E-02 2 0.68 3.68E-02 2 0.80 6.21E-02
POSTN Periostin 15 3.96 1.82E-02 18 1.26 3.79E-01 15 1.77 4.30E-02
ATP1A1 Sodium/potassium-transporting ATPase subunit alpha-1 2 1.95 2.69E-02 4 1.61 1.33E-01 2 1.74 7.77E-03
TTN Titin 15 1.38 4.03E-01 15 1.23 5.69E-01 11 1.52 3.49E-02
FBLN1 Fibulin-1 3 1.82 2.29E-03 2 1.72 1.05E-01 2 1.51 5.82E-02
Verification of findings
In view of further verifying the identified proteins analyzed via proteomics, Western Blot (WB) and/or immunohistochemistry (IHC) were performed on A. iliaca communis samples of the sysVASC patient cohort. At first, all A. iliaca communis samples were pathologically evaluated via hematoxylin eosin-stained paraffin sections. Samples were classified in 5 groups: Group A with completely thin intima; Group B with slightly thickened intima; group C with a thickened and partly strong thickened intima; Group D with a thickened intima and calcification in the intima, and Group E with a thickened intima and calcification in the intima and media (Figure 5).
Figure 5: Histopathological classification of hematoxylin and eosin-stained paraffin sections.
Using these well characterised samples several of the proteomic changes were confirmed including the shortlisted targets for interference studies (described in WP5).
1.3.3. WP3: Statistical Modelling and systems Medicine Analysis
Eagle Genomics (EAG): Part of Eagle’s role in the consortium was to combine the data from different analysis produced by the different partners. Some of these analyses enable the identification of potential new targets/genes potentially involved in the understanding of cardio-vascular disease progression. In total 9 new targets were identified by the consortium. At this stage Eagle was able to apply and refine its unique data valuation technology in order to help in the prioritisation of the most promising targets for further validation work. This technology enables the capture of subjective criteria (e.g. disease association, drugability, toxicology profile, novelty) provided by domain expert and represent in a very explicit and objective way the decision-making process. The sysVASC project has enabled Eagle to focus the application of its technology on a very concrete use case, very applicable and valuable to the pharmaceutical industry.
University of Glasgow (GLA): The stroke-prone spontaneously hypertensive (SHRSP) rat and its reference control strain, the Wistar Kyoto rat (WKY), are well-characterised models widely used in the investigation of cardiovascular disease. By comparison of the signal pathways and networks in these models it may be possible to elucidate the mechanisms which lead to hypertension and cardiac disease. We have used tissue proteomics methodologies combined with label-free protein quantitation to investigate the mechanisms underlying cardiovascular disease in SHRSP and WKY rats, with and without salt loading.
By utilising a tissue proteomic approach to both SHRSP and WKY rat aortas we have described a total of 346 significantly differentially regulated proteins. Of these, 40 were downregulated in SHRSP exposed to 1% salt while 70 proteins were upregulated under the same conditions. There were several proteins in this dataset which are known to be involved in cardiovascular disease which increases confidence in the dataset. In the WKY dataset there were 96 proteins downregulated in the presence of 1% salt and 140 proteins upregulated.
We also investigated the interactions between proteins in the datasets. We utilised a publically available database (String DB) in order to investigate these interactions. Our results highlight potentially differentially regulated pathways and signalling networks. By mapping the differentially regulated proteins in this way we are able to draw comparisons between different models of cardiovascular disease.
Helmholtz Zentrum München (HMGU): In partnership with INSERM and Biocrates, HMGU wished to determine metabolites indicative of cardiovascular risk (CVR, here using carotid intima media thickness (cIMT) as a surrogate) and diabetes in humans using the Biocrates metabolomics platform, and to determine which animal models most closely mimic these conditions. Regression analysis was used on the metabolite profiles of the study participants of the Cooperative Health Research in the Region of Augsburg (KORA) F4 population-based study (2006-2008, N=2714 for cIMT, N=3041 for diabetes) to determine differences between individuals at risk for cardiovascular disease and those showing less risk. The profiles were also compared for individuals with and without prevalent diabetes. The results were compared to those for knock-out mouse models, rat models and klotho mouse models. For both human and animals, metabolites were ranked according to their differential concentrations in the case group compared to the control group. The list rankings for each animal model were then compared to the list ranking for humans. A “top” model for CVR risk was determined, while for diabetes no model showed much similarity to humans, likely given the different types of diabetes represented (type 1 in animals, primarily type 2 in humans).
HMGU BGC was responsible for method developments and data integration of non-targeted metabolomics datasets. Non-targeted metabolomics aims at the detection and analysis of all metabolites and small molecules in a living organism so as to enable the modelling and understanding of diseases and perturbations, which is to complement strategies for metabolic control. Despite decades of analytical and technical advances in terms of sensitivity and specificity, data analytical strategies are still in their infancy in comparison with other ‘omics’. In non-targeted metabolomics analysis, typically more than 70% of signals with assigned molecular formula (e.g. C6H12O6 for glucose), but systems biology can only perform once data can be mapped to biological prior knowledge. We developed new general strategies for mass-spectrometry-based metabolomics that incorporate all signals with known molecular formula into the evaluation and interpretation of metabolomic studies. We used mass differences known to occur between metabolic m/z signals as a result of biochemical reactions to build up data-driven metabolic networks. We extend the analysis of such networks by focusing on the occurrence patterns of these known biochemical reactions, to determine how an organism builds up its body from known metabolites. This strategy – called mass difference enrichment analysis – immediately connects the metabolomic data to transcripts, proteins, disease phenotypes and other metadata. Analytical methods developed by BGC use the entirety of metabolomic datasets, therefore improve interpretability of yet unknown biochemical relationships and open a new door for systems biology to make use of metabolomics.
Mosaiques Diagnostics & Therapeutics AG (MOS): Clinical proteomics aims at developing protein/peptide biomarkers that either as stand-alone or in combination with other clinical variables will improve current medical practices. These biomarkers can aid in the management of chronic diseases i.e. diagnosis, and prognosis. However when combined with clinical parameters, high-resolution mass spectrometry proteomics generate large and complex data requiring sophisticated and robust computing statistical analyses.
Within sysVASC, we compared the effectiveness of different statistical methods to enable urinary biomarker discovery. Additionally, we investigated the performance of urinary biomarker signatures also called classifiers established by different modelling approaches for the diagnosis of coronary artery disease (CAD). In cohorts of 197 and 368 participants, respectively assessing the diagnosis of CAD and the prognosis of acute coronary artery syndromes, urine samples were analysed via capillary electrophoresis coupled to mass spectrometry (CE-MS). For biomarker discovery, different statistical methods including the wilcoxon rank sum test, t-score, cat-score, binary discriminant analysis and random forests were applied and compared. Findings depicted differences in biomarker identification depending on the statistical approach used. These differences in biomarker signatures did not translate into significant differences in the performance of diagnostic or prognostic classifiers modelled by support vector machine, diagonal discriminant analysis, linear discriminant analysis, binary discriminant analysis and random forest. Thus, findings showed that above mentioned statistical methods can be successfully used to manage proteomics data.
The analysis of mass spectrometry (MS) data to enable peptide identification with high confidence is very complex and challenging. Currently available solutions do not enable harmonising and clustering the data obtained from MS analysis. However, such harmo-nisation is essential to detect significant changes with sufficient confidence. Therefore, we developed a software solution for handling high-resolution proteomics data to enable peptide identification with high confidence. Altogether, efforts generated within this work package enabled to better understand and manage large proteomics data.
Swiss Institute of Bioinformatics (SIB): The aim of our activity was to use network-based dynamical models to identify and predict key molecular events – and possible targets for therapeutic intervention – in the evolution of cardiovascular pathologies, particularly athero¬sclerotic plaques formation.
We have chosen to adopt a logic-based modeling formalism that avoids the need for explicit determination of kinetic parameters. Our logical model is based on a prior knowledge network (PKN) that combines human and animal data from over 200 papers and includes a total of 729 nodes (including 432 proteins and 297 other entities such as metabolites and biological processes) joined by a total of 3,406 edges (regulatory interactions). Experi¬mental proteomic datasets provided by the other sysVASC partners were used to construct subnetworks from this PKN for animal and human systems; Boolean models generated from these subnetworks were then used in dynamical simulations to identify critical nodes that are potential therapeutic targets. These models generated two groups of attractors characterized by two clusters of nodes with opposing behaviors; one cluster enriched in nodes implicated in lipid metabolism (many of which are known to be negatively correlated with CVD), and a second opposing cluster enriched in nodes implicated in inflammation (many of which were positively correlated with CVD). In silico perturbation experiments were also used to identify nodes that were able to switch the network from a diseased state to a healthy state and vice versa.
University College of Dublin (UCD): UCD provided data analysis and statistical modelling in order to identify biomarkers for the prediction of risk and early diagnosis of cardiovascular disease from proteomics and metabolomics data. Biological data are usually noisy and especially data from patient samples can contain high variability. Therefore, a focus was on identifying the most suitable analysis methods. For this we compared several statistical classifier models on urinary proteomics data. Different approaches produced different biomarker patterns, albeit of similar performance. Thus, different methods can be combined to produce master patterns. Another approach focussed on dealing with missing data points in mass-spectrometry (MS) based proteomics, which is a particular issue with proteins close to the detection limit. We developed a Bayesian algorithm that exploits this knowledge and uses missing data points as a complementary source of information to the observed protein intensities in order to find differentially expressed proteins. The algorithm consistently outperformed other methods. We then applied these methods to analyse MS based proteomics data from urine, blood vessel and heart tissues. The results suggested that aberrations in extracellular matrix (ECM) and complement activation are enriched in diseased blood vessels. This indicates that remodelling of cell adhesion and blood clotting abnormalities are main contributors to the development of cardiovascular disease. In diseased hearts suffering from dilated cardiomyopathy a vastly enhance glycolysis was observed. On the other hand, in ischaemic heart disease oxidative phosphorylation and mitochondrial structure were affected.
University of Manchester (UNIMAN): The University of Manchester has worked closely with INSERM Toulouse on the problem of identifying the proteases that lead to the peptides found in urine and other body fluids. Peptides are parts of proteins formed as proteins are degraded; one mechanism for this degradation are enzymes known as proteases that cleave proteins into peptides. This cleavage is a normal part of the body's processes; however, a perturbation to these processes can result in an atypical peptide profile. Knowing which proteases are implicated in peptide production can give insights into the pathways and processes occurring within the body.
Identifying peptides and the proteins whence they came is a long studied problem. Less studied is the problem of the proteases implicated in peptide production. We have developed a Web based tool called Proteasix that tackles this question. The user provides a list of peptides and the proteins whence they come. The end of each peptide implies a cleavage point for a protease. Protasix re-constructs those cleavage sites and then matches those cleavage sites to known proteases for those sites and proteases predicted to cleave at those sites. The result is a list of proteases linked to the proteins and the peptides produced. The proteasix tool has been used successfully in analysing many data sets.
1.3.4. WP4: Evaluation of similarity of existing CVD model systems to human disease
Objective
The major objective of WP4 was to provide animal models of CVD and perform detailed molecular characterisation using omics analysis of animal tissues and body fluids and evaluate the similarity of human CVD and the selected animal models.
sysVASC’s animal models
A wide range of animal models was generated and analysed including hypertensive rats (SHRSP) and mice (angiotensin II infusion), atherosclerotic mice (Ldlr-/- and apoE-/-) with and without diabetes, an atherosclerotic plaque rupture model and a model of vascular calcification (Klotho mice). Some typical lesions in these models are shown in Figure 6.
Figure 6: Typical CVD lesions in the sysVASC’s animal model suite. A) Thickening of the vessel wall in hypertensive rats. B) Plaques in atherosclerotic mouse models which is exacerbated by diabetes. C) Calcification in the Klotho mouse model (in red). D) Increased medial thickness in mouse hypertensive model (in red) and E) Vessel segment I represents unstable/rupture-prone atherosclerotic plaques characterized by disruption of fibrous caps and luminal thrombosis in the plaque rupture model.
These models were used to produce an impressive amount of omics data that were used to describe the animal model similarity to human CVD and select appropriate CVD models for drug studies (WP5).
Urinary peptides
As a proof-of-concept study we showed that we could use urinary peptides to determine the similarity to human CVD of the atherosclerosis/diabetes mouse models. Comparing the urinary peptidome (ortholog peptides) of these mice to two previously established peptide models of human CVD (CAD238 classifier, Zimmerli et al., Mol Cell Proteomics 2008) and the ASCP75 classifier (Htun et al., PLoS One. 2017) it appears that the diabetic Ldlr-/- strain is most versatile since it might serve as a model of CAD at weeks 15 and as a model of ACSP at weeks 22, while the Apo-/- strain seems only a good model of CAD (Figure 7).
Figure 7: Abundance of ortholog urinary peptides in different atherosclerotic/diabetic strains in time.
STZ, streptozotocin (inducing diabetes).
Vessel proteins
Comparison between the mouse and rat model proteomic data with collected proteomic data from human vessels (WP2) was also performed. Mouse and rat homologues of the human proteins were retrieved manually using databases including Uniprot, the Mouse Genome Informatics database, the Rat Genome Database, and HGNC (HUGO gene nomenclature committee). In total, 109 differentially expressed proteins (between CVD cases/models and respective controls) were found in human and at least one animal model. Interestingly, multiple shortlisted candidates for IHC (described in WP2) including those further targeted via functional studies (such as KDM5D, described in WP5) were also detected changing in multiple animal models. As an example, a snapshot of results from this analysis is shown in the Table 4 below.
Table 4: Example of animal model-human comparisons at the protein level. Differentially expressed proteins in vessels of CVD cases (or models) versus respective controls were compared to identify “conserved’’ (inter-species) molecular changes associated with CVD. “yes” indicates differential expression with agreement in expression trend (up- or down- regulated) with the humans, whereas “no” indicates that differential expression was not observed.
Pathway analysis
Similar comparisons between rodent models and humans at the pathway level but based on proteomic data were also performed. In an assessment of the similarity of the sysVASC animal models with human disease based on vessel proteomes, all molecular identifiers were initially mapped to the CluSO (Clustered sequences and Orthologs) and OMAP (Ortholog mapping) databases, which are multi-species sequence and chemical compound cluster databases, in order to facilitate a comparison between animal (rat and mouse) and human data. This was followed by merging of repeated entries in each individual dataset, and alignment of all data into one discovery matrix. Simultaneously, to allow pathway mapping to existing composite large-scale metabolic and signalling cascade pathway maps, all data was mapped to mouse. All molecules (mouse orthologs) were loaded according to the comparisons in Pathvisio (v3.2.1) regardless of p-values and fold-changes and mapped to reference pathway maps using red (up-regulated) or green (down-regulated) colours (grey denotes molecules that were not present in the datasets). Large-scale maps used were derived from analyses of mouse diabetic aortic vessels, of mouse kidneys from an acute kidney injury model and ischemia reperfusion injury of the rat heart (Figure 8). However, it should be noted that such maps are a collection of reactions implicated in the diseases and do not reflect a specific disease model/state.
Figure 8: Large-scale pathway mapping. Composite map 1 (AKI, left), composite map 2 (diabetic vessel, middle), composite map 3 (Ischaemia reperfusion injury, IRA; right). Input are ALL identified molecules; red: up-regulated, green: down-regulated; grey: molecule not present in discovery matrix).
Plasma metabolites
The metabolomics data obtained with AbsoluteIDQ p180 kit and Bile Acid kit for ApoE and Ldrl KO mice with and without streptozotocine (STZ) treatment were analysed in order to identify mouse models that have closest resemblance to human diseases. Initial data analysis (Figure 9) demonstrated that the STZ treatment drastically altered the metabolic profile starting from week 15. Additionally, the ApoE KO mice with STZ treatment showed similar metabolite profile as Ldlr KO mice without treatment. A literature search was per-formed and further analysis was performed focusing on pathways and pathophysiological process reported to be altered in humans. The alterations in following pathways / patho-physiological processes were observed: hexose levels; lipid metabolism; inflammation; amino acid metabolism; arginine metabolism; acylcarnitine metabolism; beta-oxidation; bile acid metabolism.
Figure 9: Heat map of the mouse model data obtained with AbsoluteIDQ p180 kit and Bile Acid kit. Streptozotocine treatment drastically altered the metabolic profile starting from week 15.
1.3.5. WP5: Interference studies in in vivo CVD disease models
Objective
Prioritise targets, determine effects of in vivo interference on the development of CVD and define the molecular mechanism of action of interfering compounds.
Drug repurposing based on proteome human-based CVD signatures.
We used the prediction tool CMAP (Connectivity Map: http://portals.broadinstitute.org/cmap) to identify a drug potentially active in reversing the human CVD protein profile (data obtained in WP2). This in silico approach allowed us to establish a list of 53 drugs 9 of which were considered of potential clinical interest based on literature and clinician practice. One of the top compounds (for confidentiality issues called “AUFF”) was tested in vitro and in vivo for its effect in the development of CVD.
In the first step we selected the most appropriate preclinical animal model of CVD in which AUFF efficiency should be assessed. For that we compared the proteins targeted in silico by AUFF with those modified in the 11 preclinical sysVASC mouse models (described in WP4). This led to the selection of the ApoE-/- mouse model as the model where AUFF should have the largest impact.
ApoE-/- mice were treated for 10 weeks with AUFF. Interestingly, AUFF treatment was associated with a significant reduction in lipid accumulation in plaques (Figure 10).
Figure 10: AUFF treatment for 10 weeks reduced lipid accumulation in aortic roots of ApoE-/- mice.
Immunostaining analysis showed that aortic roots from AUFF-treated mice exhibited a significant reduction in collagen-III protein expression when compared to vehicle-treated mice (Figure 11). In contrast, no change in αSMA (alpha smooth muscle actin) was observed (not shown). We conclude that the anti-atherogenic activity of AUFF was associated with a reduction of fibrosis independent myofibroblast activation.
Figure 11: Collagen-III (Col-III) staining of aortic roots from ApoE-/- mice after 10 weeks treatment with AUFF (n=7) or its vehicle (n=10). *: P<0.05.
AUFF also showed a most potent effect during calcification in vitro in vascular smooth muscle cells (Figure 12). The effects of AAUFF are now being confirmed in an established mouse model of vascular calcification.
Figure 12: Effect of AUFF on in vitro calcification. A) Alizarin red staining in HAoSMCs following treatment for 11 days with control or with calcification medium (Calc.) without or with additional treatment with AUFF. B) Quantification of calcium content in HAoSMCs following treatment with AUFF. *(p<0.05) statistically significant vs. control treated HAoSMCs; †(p<0.05) statistically significant vs. HAoSMCs treated with Pi (Calc) alone.
We have thus shown that drug repurposing based on protein-based molecular signatures obtained from diseased vascular tissue can identify valid compounds with potential pro-tective effects on CVD.
Novel Targets
Bringing together the high-resolution tandem mass spectrometry analysis (LC-MS/MS) data of thoracic aortas from two atherosclerotic mice models, Ldlr-/- and ApoE-/-, and from the human central and peripheral vessels and literature, KDM5D, a specific protein involved in epigenetic regulation, was highlighted as being upregulated in CVD cases versus controls. KDM5D (also known as JARID1D or SMCY) is one of the four members of the KDM5 family of histone demethylases that specifically demethylate trimethylated and dimethylated Lys-4 of histone H3 (H3K4me3). Preliminary data (Figure 13) suggest a lower abundance of H3K4me3 in vascular tissues from patients with atherosclerosis compared to controls, potentially confirming the observed elevated expression of KDM5D in atherosclerosis. As inter-individual variability may be seen, this preliminary observation will have to be confirmed following the analysis of a higher number of clinical samples.
Figure 13: A) Western blot analysis of Trimethyl-Histone H3 (Lys4) in protein samples from i) peripheral vessels and ii) central vessels from patients with atherosclerosis and organ donors without cardiovascular background. Total Histone H3 was used as loading control. B) The quantification of the proteins from both central and peripheral vessels was performed by using the Quantity One software (BioRad) and the results were normalized to total H3.
Given the observed up-regulation of KDM5D and the downregulation of its substrate H3K4me3 in disease versus control, we sought to evaluate the effect of KDM5D inhibition in HUVEC cells. Inhibition of KDM5 resulted in an increase in the levels of the trimethylated state of lys4 of histone 3 (H3K4me3) as supported by Western Blot analysis (data not shown). Interestingly, the KDM5 inhibition also resulted in a significant reduction, by at least 40% in both the proliferation and migration capabilities, with a further reduction by at least 20% in the number of nodes, junctions or segments developed by the HUVEC cells. As an example, representative images from the investigation of the tube forming ability of the cells are shown below (Figure 14).
Figure 14: Representative images from the study of impact of KDM5D inhibition on the tube-forming ability of HUVEC cells. Three independent experiments were performed in duplicate each time.
Collectively, these data suggest a role for KDM5 histone demethylases in the development of atherosclerosis likely via affecting the H3K4 methylation. Further experiments investigating the impact of the KDM5 inhibitor on plaque formation are required.
Another target that derived from comparison of omics datasets across species and different types of vascular diseases as described is sirtuin-1. Sirtuins are critically involved in the regulation of mitochondrial processes and have previously been implicated in ageing, inflammation and apoptosis – all of which are features of vascular diseases. The widely used antioxidant resveratrol is known to be an unspecific activator of sirtuins. This has prompted us to study the effect of resveratrol on vascular function and structure in the SHRSP, a well established rat model of hypertension that has also informed the proteomic data leading to discovery of differentially regulated sirtuin-1 in our datasets.
Animals were exposed to 10 weeks of resveratrol treatment (130 mg/kg/day) from 5 weeks of age and underwent weekly blood pressure measurement. Wire and pressure myography have been conducted on mesenteric arteries. Our data suggest that the Sirt1 activator resveratrol has no statistically significant impact on blood pressure in the SHRSP. However, on a structural level, resveratrol-treated animals show an improved distensibility profile (stress - strain relationship) when compared to controls, which suggests improved vascular remodelling (i.e. reduced vascular stiffness in the resveratrol treated rats) (Figure 15).
Figure 15: Pressure myography in mesenteric arteries from resveratrol and vehicle treated SHRSP rats. A) external diameter, B) internal diameter, C) lumen diameter, D) cross sectional area, E) wall stress and F) vessel distensibility.
1.4. WP6: IP Management and Exploitation
Coronary artery disease (CAD) is a major public health concern in developed countries. Although a significant improvement in the management of CAD was observed, the prevalence of CAD is still on the rise due to demographic transition and increased survival rates. The aetiological diversity of CAD makes it challenging to fully decipher complex pathological processes based on single biomarkers. It is therefore obvious that innovative methods that will improve the management of CAD are urgently needed. Despite the socio-economic burden caused by CAD, the development of innovative therapeutic compounds has been moderate due to a limited understanding of primary mechanisms associated with the disease.
The scientific aims of sysVASC were to address current obstacles in the field of CAD including:
• Limited understanding of molecular mechanisms associated with CAD.
• Limited access to human samples.
• Lack of suitable mouse models for therapeutic interventions.
• Lack of efficient drugs targeting causative molecular mechanisms of the pathology.
Beyond these aims, sysVASC additionally focused on exploring the translation of research findings into health care applications. To achieve an effective translation and based upon high level of interaction between partners, a work package was dedicated to intellectual property management and exploitation. To translate scientific discoveries into clinical implementation, several objectives were defined at the beginning of the project including:
• Design of activities aiming at the protection of intellectual property rights.
• Direction of identified targets towards effective exploitation by appropriate third parties.
• Setup and update a database of state-of-the art in cardiovascular disease drug development and relevant (pre)clinical trials.
• Efficient exploitation of results during and beyond the lifetime of the project.
During the project, efforts were generated by partners involved in this work package to achieve the above mentioned goals. This led to the exploitation of sysVASC findings during and even beyond the duration of the project. At first, a strategy to optimise exploitation was setup and the main principles for the management of the intellectual property rights were as follow:
• Results were openly disseminated in parallel with appropriate protection of intellectual property to encourage the promotion of scientific results during the project. In case of publication, reasonable notification and deferral procedures to ensure intellectual property protection were set in place and followed.
• Background intellectual property remained the property of the partner (s) that brought it to the project. Access rights to Background intellectual property was granted on the basis of: 1) project use (‘for implementation’) and 2) exploitation (‘for use’) to another partner needing to exploit the data in accordance to agreement between concerned partners. As an example of Background intellectual property, Mosaiques Diagnostics GmbH had access to the intellectual property rights on biomarkers for cardiovascular disease prior to the project.
• Results of sysVASC (Foreground intellectual property) were owned by the partner(s) responsible for generating the findings of interest.
Generally, issues relating to intellectual property rights were managed in such a way that exploitation decisions were made to serve the needs and goals of the partners.
During sysVASC, novel therapeutic interventions to tackle CAD were identified. As a result, potential targets for intervention in CAD were identified. To optimise the exploitation of the findings, a patent search was performed with the compounds of interest to further investigate the novelty of the results. As translation of results in industrial application was envisaged, this was a very curial point that was carefully explored in the exploitation strategy. The patent search that was made, covered biomarkers and therapeutic targets described in the context of cardiovascular diseases. Specialised websites were used to screen for European as well as international patents that were either granted and/or published. Additionally, patent search and frequent updates enabled to ensure the potential financial value of identified therapeutic targets. Results of the patent search were helpful to redefine when necessary a relevant context-of-use of a targeted molecule or compound.
The aim of sysVASC was to exploit this unique know-how and transform it either into highly innovative products or services to generate business opportunities. The unique sysVASC know-how or approach (Figure 16) offers solutions to assist in a more successful way drug development which is of high relevance to pharmaceutical industries.
Figure 16: The sysVASC approach. Datasets and related clinical data were combined with existing knowledge on CAD from literature and omics data repositories. Data were analysed using appropriate bioinformatics and statistics tools to generate an initial model of relevant molecular and pathway changes in CAD. Key components of these pathways were verified in appropriate human samples. In parallel, animal models were investigated for molecular homology, enabling to choose the model system best reflecting the human disease. Intervention studies in this model enabled assessing its relevance to disease, guiding the development of drugs directly addressing molecular causes of CAD.
As result, pharmaceutical companies represented the targeted market group and were considered as customers of choice. Benefits of the know-how for pharmaceutical companies included:
• Access to unique human vessel tissue samples to identify adjacent proteomic changes relevant in CAD.
• Access to the unique surgical tandem stenosis mouse model reflecting all stages of human atherosclerosis.
• Access to a unique database of urine samples to help achieve a systemic assessment of CAD-associated changes and their validation.
• Access to comprehensive proteomic analysis using cutting-edge technologies for a better and an in-depth protein characterisation.
• Access to proprietary software solutions to evaluate proteomic data.
• Access to proprietary tools to facilitate the selection of animal models best displaying the human disease-phenotype based on proteomics.
• Access to proteomics and bioinformatics tools to facilitate drug target identification.
• Access to a unique proteomic expertise breaking through the norms of protein identification and characterisation by further exploiting the information to identify potential drug targets and select appropriate animal models for intervention studies.
Furthermore, to achieve effective dissemination in the above mentioned market within the duration of the project, additional efforts led to the development of a marketing plan.
To generate business opportunities with pharmaceutical companies, a dissemination database was developed and updated. The database included collected information on current state-of-the-art in cardiovascular disease drug development and relevant clinical trials. Additionally, other relevant information including the name of the principal investigator of a trial, and/or the name of an employee working in the area of cardiovascular disease in a pharmaceutical company along with a relevant email were included. Using the email addresses included in the database, people of interest were contacted to initiate business opportunities. The dissemination database was a key tool to promote exploitation of results because it helped reach the targeted audience.
To further contribute to the successful exploitation of the project, a proposal collecting key findings and highlighting the unique approach used to improve drug development was generated at the end of project. The proposal summarising results generated within sysVASC was used as a basis for contacting pharmaceutical industries and establishing new collaboration to further validate novel therapeutic targets in clinical studies. Using contact details collected in the dissemination database, the proposal was disseminated to people of interest. As this process is ongoing, potential business opportunities emanating from this initiative can only be evaluated after the project. However in the meantime, a negotiation strategy was developed to ease conversation with pharmaceutical companies. Briefly, the negotiation strategy will include preparation, discussion, proposing and bargaining phases that will help secure a successful collaboration.
Altogether, the sysVASC project generated exploitable intellectual property proposing novel strategies (i.e. therapies) to improve the management of CAD. Exploitation of these resources enabled to initiate business opportunities.
Potential Impact:
1.5. Potential Impact
Within sysVASC, we implemented a systems biology approach to gain knowledge on the molecular pathways underlying CVD. The multidisciplinary effort benefited from (i) excellent preclinical and clinical research centres, (ii) cutting edge omics technological platforms and (iii) unique bioinformatics and statistics expertise. With the broad, essentially hypothesis-free approach, sysVASC was designed to become a showcase for the application of systems biology in clinical research and development. The integration of previously disparate data sources generated a unique comprehensive expandable resource for identifying and validating novel key molecular targets for treatment, thus tackling the lack of interventions specific for CVD and having major scientific, societal and economic impact. Identified targets for CVD treatment were advanced towards translation into novel therapeutic approaches and may even result in additional products such as biomarkers for diagnostic/prognostic purposes, thereby relieving a major burden to patients and health care systems and promoting CVD treatment towards personalised patient management and individualised targeted therapy.
1.4.1. Potential impact of the results of WP1
WP1 has developed a number of tools that facilitate the integration of multi-omics data across different datasets in clinical cohorts. The tools that have been developed and described above are available to the scientific community. Two resources should be mentioned in particular in this context.
First, the CVDKB is accessible to researchers worldwide (http://cvdkb.org/) and supports data analysis and integration. The experiments span over different biological levels (genes or proteins), different techniques, different species (mouse or human), and different sample types (blood or coronary artery). Experimental data derive from published articles in Pubmed, all related to coronary artery disease. The data have been extracted manually from the figures or from the supplementary data when possible. Microarray data available in the gene expression omnibus (GEO) were reanalysed in house using standard R scripts (GEO2R). Altogether, 34 publications met the inclusion criteria and 26 were excluded.
Second, e[ADS] (developed by Eagle Genomics) has several impacts for clients in academia and industry:
• Improved time and ability to insight through improved productivity of researchers.
• Faster identification and generation of successful product innovation and insight, leading to reduced time to market.
• Competitive advantage through cutting edge research, driving innovation in the product pipeline.
• Efficient compliant, documented, evidenced, and enforced data sharing process; effective (time & cost) collaboration between vendors and internal teams.
• Systematic data sharing and re-use, driving faster unique cross-functional insights and reducing costs.
• Rapid seamless integration of diverse data sets for scientific analysis.
1.4.2. Potential impact of the results of WP2
• Generation of an invaluable resource of well characterized clinical samples and comprehensive molecular data to support further integrative studies on CVD in the future.
• Lists of verified proteomic changes in human CVD of high biological interest, whose further investigation as potential therapeutic targets is warranted.
• Optimized, reproducible procedures for sample preparation and workflows for proteomics data analysis of starting material of limited amount, disseminated to the community to increase comprehensiveness and data reproducibility of similar pre- and clinical proteomics studies in the future.
1.4.3. Potential impact of the results of WP3
Eagle show cases this data valuation use case at any customer opportunity to demonstrate its strategic potential in target portfolio management and has already opened up other opportunity where the prioritisation of other “entities of interest”, e.g. best patients for stratification, best indications for drug repurposing or best natural ingredients for new skin care cream formulation.
HMGU: Metabolomics represents the closest omics level to clinical phenotype and can inform on both products of disease or possible drug targets. The qualitative and quantitative knowledge of the metabolic similarities between animal models and human cardiovascular risk can allow us to better plan future cardiovascular disease studies using the most appropriate model, either targeting a small collection of specific metabolites or utilizing the metabolomics profile as a whole.
MOS: Proteins regulate all biological functions and allow an optimal representation of physiological, environmental and pathological variations occurring over time. Thus, proteome analysis is a suitable approach providing necessary information on molecular pathological processes and enabling a better disease characterisation. Recent advances in proteomics and the development of robust tools (mass spectrometry (MS)-based) for an in-depth characterisation of proteins now enable comprehensive investigation of proteomic alterations in complex biological samples. However, identification of peptides with high confidence remains a challenge. In this work package, we have developed a software solution to improve peptide identification. Therefore, accurate protein identification will contribute to a deeper understanding of molecular mechanisms associated with diseases and improve disease management. As a result, the impact of the software solution will move beyond the scope of sysVASC.
SIB: Our computational simulations have allowed us to identify a list of potential therapeutic targets that can be validated experimentally. A promising target that was identified in all analyses using models built using proteomic data from animal and human tissues was the protein SIRT1, which has previously been shown to have a protective effect on cardiovascular disease and which in our simulations appears as a promising therapeutic target with the ability to reverse the disease phenotype when overexpressed/activated.
UCD: These studies have provided new analysis methods for proteomics data from human body fluids and tissues. The results suggest new biomarkers for the diagnosis of cardiovascular disease and different forms of cardiomyopathies.
1.4.4. Potential impact of the results of WP4
Comparison of human disease and preclinical animal models of CVD at the molecular level will lead to the:
• definition at the molecular level of the most appropriate models representing certain aspects of human CVD for subsequent molecular and interference studies;
• development of “humanized” readouts in mice that are supposed to significantly speed-up the translation from the preclinical to the clinical phase since we anticipate that successful drugs as monitored based on these humanized readouts will remain successful candidates during the move from animals to humans;
• towards reduced use of animals models in line with the 3Rs European policy
o For drug testing, the in silico selection of the model most closely mimicking certain aspects of human disease among a panel of available models of CVD avoids testing blindly all available CVD models (Replacement and Reduction).
o In addition, the development of non- or minimally-invasive (humanized) readouts will lead to the definition of alternative end-points (Replacement) and will allow for longitudinal follow-up during drug-testing without animal sacrifice (Reduction).
1.4.5. Potential impact of the results of WP5
The potential impact of WP5 is threefold:
• Conceptual: show-casing that the pipeline as developed in sysVASC is feasible. The combination of the sysVASC human omics data, bioinformatics pipeline and the sysVASC’s animal CVD model suite has allowed the identification of new drugs and targets. Such pipeline is applicable to any disease where omics data can be obtained.
• Identification of novel CVD targets: this pipeline focusing on the molecular mechanism of CVD has generated a number of original and very diverse targets with a focus on calcification, ECM and epigenetic modifications. This shows that detailed insight in the molecular mechanism of disease obtained by omics analysis allows the identification of very diverse novel targets of disease.
• Drug repurposing speeding up in vivo validation: in silico analysis of reversal of CVD signatures by a large panel of drug signatures generates compounds that can be readily tested in vivo.
1.4.6. Potential impact of the results of WP6
The sysVASC project aimed at providing innovative approaches to facilitate drug discovery by identifying novel therapeutic targets. The novelty of the proposed approach relied on the fact that “omics” platforms combined with literature mining were used to identify drug targets after application of robust bioinformatics tools.
Several pharmaceutical industries are actively involved in the development of novel drugs to improve the management of disease. These industries include Novartis, Pfizer, Roche, Sanofi, Merck & Co., and many others all wanting to develop the most effective drug at a low cost. The sysVASC approach makes it possible for pharmaceutical industries to achieve these goals. This can be achieved using the sysVASC approach by identifying CAD-specific biomarker signatures in tissue, blood and urine samples. Together with a thorough investigation of the literature and aided by robust bioinformatics tools, relevant molecular pathways can be investigated leading to the identification of therapeutic targets. The unique biomarker signatures can in parallel also enable the non-invasive selection of animal models and drug monitoring and a reliable patient stratification. Contacting pharmaceutical companies has the potential impact to generate business opportunities and influence the field of drug discovery in cardiovascular disease at large.
Since the sysVASC approach regrouped a multitude of technologies, this enabled a rapid identification of drug targets. Moreover, with the access to animal models, identified targets could be validated. Hence, the approach was time-efficient and cost-effective.
The potential economic benefits for pharmaceutical industries in using the sysVASC approach will thus include:
• The development of more effective drugs. Due to a comprehensive “omics” analysis provided by the sysVASC approach unique disease signatures can be generated, novel drug therapies specific for the treatment of CAD can thus be identified. This will significantly reduce the costs as compounds to be tested in a clinical trial will be of high confidence. For instance, developing a compound with high confidence would have prevented some financial losses already made by pharmaceutical companies in an attempt to improve the management of cardiovascular diseases.
• Better selection of animal models and non-invasive monitoring. The sysVASC approach has the unique advantage to enable the selection of suitable animal models best mimicking the human disease. Furthermore, the efficacy of a therapy can also be assessed non-invasively in the selected animals. This will contribute to the increase rate of success of preclinical studies ultimately leading to clinical studies with high confidence.
Overall the sysVASC approach has the potential impact of generating business opportunities with the pharmaceutical industry. As the sysVASC approach proposes new strategies to improve the management of CAD, it can additionally help reduce the current economic burden caused by CAD.
1.6. Main dissemination activities and exploitation of results
Professional project management and dissemination of results significantly contributed to the successful collaboration of the sysVASC beneficiaries. The Coordinator encouraged all participants to actively disseminate their project aims and results to colleagues, stakeholders and the wider public.
At the beginning of the project, a logo and a corporate identity for the project was created in order to identify the project. It is used together with the Grant Agreement number HEALTH.2013.2.4.2-1 603288 and the FP7 logo (until use of the FP7 logo was officially discontinued) on all printed and electronic material for the public and for any other official contact.
On the web, public awareness of sysVASC was addressed by the project website www.sysvasc.eu. The website content was regularly updated and includes key background information about the project and its beneficiaries. All content was available in two languages: English and German. Additionally, project fact sheets are available in 4 languages for external stakeholders allowing information to be obtained at a glance. The news and media section provided visitors with the latest publications, press releases and dissemination activities of sysVASC, as well as the sysVASC postcard quiz, which was released at the MUG 10-year anniversary, an event reaching 3000 people, to generate greater exposure for the project and increased impact.
The centrepiece of the last website update was the sysVASC 3D video, which clearly presented the background to the project and highlighted some of its successful outcomes.
The Consortium also took measures to provide sysVASC exposure to the public through conventional PR media work and new media, which complemented the public access website, particularly for those that do not have internet access.
The scientific community has a vested interest in keeping up with sysVASC progress and results to build on the body of knowledge that currently exists. This was achieved in part through conventional scientific dissemination means, i.e. peer reviewed publications and presentations at international conferences. 40 peer-reviewed publications were published during the project funding period as part of the project and partners have presented sysVASC work world-wide 20 times at international scientific conferences. These publications also included a significant number with contributions from multiple partners, emphasising the impressive level of cross-consortium cooperation, which the sysVASC project has demonstrated. A number of additional publications are at the preparation or review stage and are expected to be published in the near future.
Two dissemination highlights that occurred during the first half of the project were the Medical university Graz 10th anniversary and ‘Destination LABO’. As part of its 10-year anniversary celebrations the MUG introduced the sysVASC project to the public. 3000 visitors attended the event, held on the 17th October 2014 in the main square of Graz. At the ‘Destination LABO’ event the INSERM’s renal fibrosis laboratory in Toulouse opened their doors to students from primary and secondary schools to present current work regarding chronic kidney disease and its association with cardiovascular disease. A special emphasis was placed on the sysVASC project during the event.
List of Websites:
Name of the scientific representative of the project's coordinator and organisation:
Prof. Dr. Burkert Pieske
Charité – Universitätsmedizin Berlin
Augustenburger Platz 1
13353 Berlin | Germany
Phone: +49 30 450 553702
Fax: +49 30 450 7553 702
E-mail: burkert.pieske@charite.de
Project website address: http://www.sysvasc.eu/
1.1. Executive Summary
“Systems Biology to Identify Molecular Targets for Vascular Disease Treatment” (sysVASC) was a collaborative project based on established multidisciplinary European research networks, including specialists in pre-clinical and clinical research, omics technologies, and systems biology from research intensive SMEs and academia. sysVASC was funded through the European Union’s Seventh Framework Programme (FP7- Collaborative Project (CP) - Small or medium-scale focused research project (STREP)) and was officially launched in February 2014 and ran for 48 months. The project website can be found at http://www.sysvasc.eu/.
Cardiovascular disease (CVD) is the leading cause of mortality and morbidity in Europe and worldwide. Asymptomatic vascular damage accumulates for years before patients are identified and subjected to therapeutic measures. The limited knowledge on early vascular disease pathophysiology is reflected in the lack of therapeutic options. Prognostic biomarkers still remain widely elusive and personalised risk assessment relies on rough estimates based on traditional risk factors. The underlying pathophysiological changes remain incompletely understood. The main clinical needs faced in CVD are: (i) a better understanding of the key pathogenic factors involved in the progression of CVD independent of aetiology, (ii) the diagnosis of asymptomatic disease and (iii) the prediction and prevention of cardiovascular events.
sysVASC aimed to overcome this limitation by identifying pathophysiological mechanisms and key molecules responsible for onset and progression of CVD and validation of their potential to serve as molecular targets for therapeutic intervention. The sysVASC consortium gathering a wide range of experts in the field of different “omics” technologies, systems biology, statistics, cardiovascular diseases and preclinical studies provided a unique know-how necessary to break through current challenges in the field of CVD. As a result, the molecular phenotype of progressive CVD was identified using high resolution “omics” data in both humans and animal models. We collected multilevel data of 1949 individuals with CVD that in a systems biology analysis integrating clinical features, tissue and body fluid multi-omics data and extensive literature data led to the identification of a number of original targets in CVD focussing on calcification, extracellular matrix and epigenetic modifications. In parallel a wide range of animal CVD models were characterised to the same molecular level as to determine which of the animal models represent optimally human CVD. These models were used in experiments aiming to interfere with sysVASC targets leading to confirmation of the relevance of these targets in CVD and their suitability for further drug-development.
Project Context and Objectives:
1.2. Project Context and Objectives
1.2.1 Project context
Cardiovascular disease (CVD) is the leading cause of mortality and morbidity in Europe and worldwide. The term CVD comprises a multitude of clinical conditions and different degrees of severity. The pathophysiology of CVD is complex and multifactorial and the limited knowledge on early vascular disease pathophysiology is reflected in the lack of therapeutic options. Before patients are diagnosed and treated, asymptomatic vascular damage can accumulate for years. Although early pathophysiological changes are at least in part reversible, their diagnosis remains challenging. They are clinically asymptomatic and require specialised equipment to characterise the vascular phenotype, and a universally accepted method to assess vascular function and structure in the clinical context is not available. The main clinical needs faced in CVD are: a) a better understanding of the key pathogenic factors involved in the progression of CVD independent of aetiology, b) the diagnosis of asymptomatic disease and c) the prediction and prevention of cardiovascular events. Prognostic molecular biomarkers remain widely elusive and clinical practice currently relies on rough estimates based on traditional risk factors. Consequently, current preventative measures are of general nature and aim at these risk factors without taking their relative contribution to an individual's pathophysiology into account. It is presumed that a clearer understanding of the pathophysiology of CVD will promote personalised risk assessment and targeted preventative and therapeutic measures.
The sysVASC Consortium aimed to overcome this limitation by mounting a comprehensive systems medicine approach to elucidate pathological mechanisms, which will yield molecular targets for therapeutic intervention. The sysVASC Consortium compiled the multi-disciplinary skills and unique expertise necessary to tackle this ambitious task. We integrated high quality, high resolution molecular omics data from clinical studies in order to comprehensively describe the molecular phenotype of progressive CVD. Statistical modelling identified targets for specific interventions. These have been initially validated in models best mimicking key pathogenic processes in human disease.
1.2.2 Project Objectives
The main objective of the sysVASC project was to identify key factors involved in the pathogenesis and progression of coronary artery disease (CAD) and to identify molecular targets for the treatment of CVD. The project had 5 more specific objectives that have been addressed throughout the lifetime of sysVASC.
Objective 1: To identify the key pathogenic factors involved in the progression of CAD and leading to coronary events.
The sysVASC consortium decided to focus its research on macrovascular disease. For literature-based work CAD has been chosen as model disease due to the very clear definition of the two key clinical events: myocardial infarction and death from a cardiovascular event. In experiments in experimental models and human tissue samples the consortium benefited from detailed histological analyses and/or functional data on vascular phenotypes. In large cohort data we applied a specifically developed staging model in order to estimate severity of vascular disease.
Objective 2: To study the molecular features of CAD independently of aetiology.
Pathogenic factors including hypertension, diabetes, smoking and dyslipidaemia trigger different processes in the initiation of vascular diseases. In more advanced, yet still clinically asymptomatic stages, the progression is in part independent of the original aetiology. The sysVASC consortium collected therefore well phenotyped clinical samples that were used for in depth multi-omics molecular analysis (transcriptomics, proteomics/peptidomics, targeted and untargeted metabolomics) and for the integration of clinical, pathological and molecular phenotypes towards understanding the CVD pathophysiology.
Objective 3: To apply systems medicine approaches to define the key nodes of pathogenic networks involved in the progression of CVD.
Complex multifactorial and chronic pathologies such as CVD cannot be described on a single molecule level. Instead, a comprehensive analysis needs to uncover nodes within a complex framework of interactions, which maintain disease mechanisms. sysVASC thus developed an ambitious systems medicine approach to uncover molecular pathways involved in the progression of macrovascular disease from early stages to clinical events. These key factors and pathways identified in human studies were then targeted in the most appropriate experimental models to prove their role in disease progression and as targets for preventative and therapeutic approaches.
Objective 4: To identify animal models mimicking specific key elements of human CVD.
In parallel to human data, the sysVASC consortium carried out an in-depth characterisation of CVD development in animal models representing the key risk factors, including hypertension, diabetes, hypercholesterolaemia, hypercalcemia and combinations thereof. This suite of animal CVD models were characterised to the same level of detail as human samples by applying proteomics, peptidomics and metabolomics to the animal body fluids and tissue, enabling precise identification of common features between animal models and humans. This led to the definition of the most appropriate models representing certain aspects of human CVD for subsequent molecular and interference studies. A number of these animal models combined with the massive omics data also served as starting points of novel CVD targets.
Objective 5: To validate new therapeutic targets hampering the progression of CVD in the most appropriate animal models.
The systems biology analysis components of sysVASC lead to the identification of a number of novel CVD targets and their associated compounds for interference with their activity. Other computational approaches based on the sysVASC omics data lead to repurposing of compounds. Using animal models best representing human CVD, we carried in vivo interference studies using these compounds to determine their suitability as candidates for drug development. For a number of compounds we obtained in vitro data supporting their potential as targets in CVD. Very diverse and novel CVD targets were validated with a specific focus on calcification, ECM and epigenetic modifications
Project Results:
1.3. The sysVASC Work Packages and Major Scientific Results
sysVASC was organised in five research based Work Packages (WP1-5) and three Work Packages on exploitation, dissemination and project management (WP6-8) (Figure 1).
Figure 1: sysVASC Work Packages and their interaction.
WP1 focussed on (i) the establishment and maintenance of a central data repository to deposit and integrate all available phenotypic and omics data; (ii) the development of a new CVD knowledge base containing information extracted from existing databases and literature searches; (iii) systemic analysis of (a) existing data and (b) new data on human samples generated within the consortium.
The main task of WP2 was (i) the generation of human CVD omics datasets focussing on human vascular tissue and (ii) the verification of changes of novel targets predicted by the in silico models.
WP3 used the existing and newly generated CVD data for (i) statistical modelling; (ii) identification of association networks and key nodes; (iii) advanced system biology modelling including inter-species comparison.
Within WP4 existing animal models of the key risk factors for CVD (hypertension, hypercholesterolemia, diabetes, and hypercalcemia) have been (i) molecularly characterised using omics technologies; (ii) evaluated regarding their similarity to the human disease.
The nature of the compounds to be tested within WP5 was in the beginning unknown, as this depended on the data from WP3/WP4. WP5 (i) tested the novel targets identified within the other WPs in interference studies in CVD models; and (ii) defined the molecular mechanism of action of those interfering compounds.
Key achievements of the Work Packages are highlighted in the following section.
1.3.1. WP1: Existing multi-source data integration from human disease
The sysVASC consortium benefited from human datasets that were brought into the project by consortium partners. A number of large-scale cohort data such as the “Kooperative Gesundheitsforschung in der Region Augsburg" (Cooperative Health Research in the Augsburg Region; KORA) cohort with associated omics datasets were available for in-depth analysis in WP1 and generation of additional data in other work packages. Not all of these clinical datasets have been generated primarily to study vascular diseases. Moreover, specific vascular phenotypes such as data on endothelial function, vascular stiffness, atherosclerosis; imaging data including coronary angiography; and tissue samples for detailed histological analysis of vascular damage were not consistently available in these cohorts. This is a common feature of such large datasets and the consortium was well aware of the challenges related to discrepancies between the availability of precise molecular phenotypes and less detailed and particularly, unstandardized phenotypic characterisation across the clinical datasets. These challenges extend further to literature data that may contain relevant information but cannot immediately be integrated with other data due to lack of standardisation of phenotypes.
WP1 therefore employed a number of strategies to facilitate data integration across clinical datasets that were directly available to the consortium and datasets that were obtained from literature and publicly available database searches. The key achievements of WP1 in this respect are:
• Development of a classification system of severity of vascular disease based on commonly available risk factors and clinical features.
• Development of a consortium-wide data repository to facilitate integration of data across human studies and also across other work packages. From here a more sophisticated platform has been developed to support the wider research community in their data integration efforts.
• Development of an integrated database featuring clinical and molecular data across human cohorts.
• Development of an online database of molecular features associated with vascular disease based on available literature data.
The classification system for vascular diseases has been developed to grade vascular disease severity in cohorts that do not have specific vascular phenotypic information available and to allow comparison between cohorts that employed different approaches to characterise vascular disease. It is based on other grading systems that are in clinical use, e.g. for the severity of other cardiovascular disease (Figure 2).
Figure 2: Classification system for severity of vascular diseases.
The system is based on 5 stages from A (not at risk for and no evidence of established vascular disease) to E (evidence of structural vascular disease). For practical reasons it is a combination of risk assessment and assessment of established disease and therefore remains an element of uncertainty. The grading system has therefore been used throughout the duration of the project for analysis of large cohorts where inherent imprecision can be compensated by large numbers and for screening experiments where a first selection of samples was based on the grading, followed by in-depth analysis of specific vascular phenotypes and/or histology. Where required the higher grades (D/E) were combined and contrasted against lower grades (A/B).
In parallel to this clinical grading system the consortium has made use of specific vascular phenotypes including quantitative coronary angiography, pulse wave velocity and carotid intima-media thickness. A parallel histological grading system has also been developed and applied where tissue samples were available: A: completely thin intima; B: intima slightly thickened; C: intima mostly thickened and partly strong thickened; D: thickened intima + calcification in intima; E: thickened intima + calcification in media. This system played a more crucial role for other work packages, in particular for WP2 (Figure 3).
Figure 3: Samples of common iliac arteries from a patient with histological grade A (left panel) and a patient with histological grade E (right panel) vascular disease.
In order to make all project data accessible to project participants a dedicated data repository has been developed. Secure data sharing between consortium members has been facilitated by means of an sFTP (FTP over ssh) managed server. Information security compliance is achieved by having SSL certificate encryption, regular backups and 24/7 monitoring by the server provider. Storage capability is suitable to the expected size of the project. The server has been administered by Eagle Genomics and users accessed the site via password control that has been communicated to a restricted number of consortium partners. The sFTP site has been available for use by all sysVASC partners and access to the data will be maintained beyond the funding period.
Based on the experience within sysVASC Eagle Genomics have been developing the e[automateddatascientst] (e[ADS]) platform. This platform aims to facilitate data integration and exploration in life sciences. It is a next generation conversational learning platform for data integration and discovery. The platform has 5 main modules that adhere to concepts concerning data cataloguing, data curation, data analysis, data valuation and genome visualisation. It is populated with data imported from various data sources that cover many research areas: genome, genes and transcripts from Ensembl; proteins from UniProt; pathways from Reactome; SNPs from dbSNP; association with disease from OpenTargets; pharmacology from DrugBank; and scientific literature from PubMed and EuroPMC. Data extracted from these public resources allow navigation between entities of interest, showing the evidence based relationships between each entity. This platform is constantly being developed in response to customer requirements, identifying new data sources and extending the data model with new entities. While the indication of interest for sysVASC is cardiovascular disease, the data model can be adapted for any disease area and is therefore attractive to many potential customers by delivering a range of business benefits such as improved productivity and collaboration which combined speed up time to insight.
In order to directly explore the available omics datasets across a number of clinical cohorts and within cohorts the sysVASC consortium required an additional solution that brings data together in an easily accessible format. Led by project partner Mosaiques Diagnostics the consortium has developed a database integrating resources from all partners within sysVASC that can be accessed online (http://sysvasc.mosaiques.de). Table 1 provides an overview of the large amount of data that were added to the database through regular updates.
Whilst these efforts focussed on data that are immediately available to the consortium there have been further efforts to bring literature data and data from public repositories into the sysVASC consortium. A new Cardiovascular Disease knowledge base (sysVASC knowledge base; http://cvdkb.org/) containing information extracted from existing databases and literature searches has therefore been developed based on an existing integrated knowledge base for kidney diseases (www.kupkb.org). The construction of the CVDKB involved the conversion of various external databases and required collection of experimental findings from high-throughput experiments investigating gene, protein and metabolite in the context of coronary artery disease. A decision has been made to focus on stable coronary artery disease in order to have a precise definition of a vascular disease phenotype available for implementation of the knowledge base. Additional features can be added in the future.
Table 1: Clinical and omics features contained in the sysVASC integrated database.
Features in the database Type of data Update 1 Update 2 Update 3 Update 4 Update 5 Update 6
DB_Patient Clinical 616 1,658 2,274 2,274 2,274 2,320
DB_Sample Clinical 1,074 2,111 3,184 3,184 3,184 3,184
DB_tbEvaluation Clinical 1,766 1,890 2,506 2,506 2,506 2,506
DB_OrigSample Clinical 631 1,659 2,289 2,289 2,289 2,477
DB_Analyses Clinical 1,028 2,239 3,267 3,267 3,267 3,267
DB_Diagnoses Clinical 845 1,746 2,591 2,591 2,591 3,739
DB_Parameters (biochemical) Clinical 3,579 15,366 18,945 18,945 18,945 21,872
DB_Drugs Clinical 0 0 0 0 1,502 1502
DB_CE_MS “Omics” 3,460,004 3,460,004 3,460,004 3,460,004 14,074,069 14,074,069
DB_LC_MS “Omics” 69,890 69,890 69,890 69,890 69,890 298,085
DB_Analyte_List “Omics” 0 0 0 0 0 451
DB_Analyte_Data “Omics” 0 0 0 0 0 525
The datasets present in the CVDKB can be divided into three groups (Figure 4): the domain ontologies, which add semantic annotation on top, to the experimental data to be queried; the omics database that helps linking the different biological layers; and the experimental data associated with coronary heart disease.
Figure 4: Schematic structure of CVDKB showing experimental data connected to background knowledge and annotated with the domain ontologies. Ontologies capture the relationships between entities and express these relationships in a language that can be interpreted by a computer. Ontologies are needed in order to provide a common vocabulary for describing cardiovascular anatomy, cells and diseases.
1.3.2. WP2: Generation of additional data and verification of predicted targets in human samples
The main objectives of WP2 included:
1) Collection of well phenotyped clinical samples in accordance to latest ethics regulations that can serve as an invaluable resource for in depth multi-omics molecular analysis and integration of clinical, pathological and molecular phenotypes towards understanding CVD pathophysiology.
2) Generation of multi-level omics datasets on human CVD to complement existing knowledge. This analysis included studies at all genome, transcriptome, proteome/peptidome and metabolome levels on tissue, plasma and urine, as applicable.
3) Verify differential expression of selected molecular features via a combination of tools including cross-omics comparisons, extensive literature mining as well as immunoassays.
1. Collection of clinical samples
By the end of sysVASC, 138 patients were included in the sysVASC study, with the number of vascular tissue samples (i.e. different vessels) per patient varying. Clinical data from patients are organised in a database structure to support statistical computation. The entire sample cohort is stored at CHA.
The Table below (Table 2) provides an overview of the samples contained in the collection until M48. The stages (A-E) are in accordance with the classification system established in WP1.
Table 2: Human samples available from de novo collection at MUG by the end of Period 3
Collection of omics datasets
Multiple new omics datasets were collected in the course of sysVASC. These included:
• Whole-exome-sequencing and expression profiling by RNA sequencing from a total of 33 tissue (n = 32 patients) samples.
• High resolution proteomics data from 43 plasma, 108 serum, and 70 vessel (n = 59 patients) samples.
• Urine peptidomics datasets from approximately 1000 samples (samples from the French and euRopean Outcome reGistry in Intensive Care Units (FROG-ICU) study were also included in the analysis).
• Targeted metabolomics datasets from 126 plasma, 108 urine and 66 tissue (n = 55 patients) samples. Untargeted metabolomics using FT-ICR from 149 tissue (n = 42 patients) samples, 43 matched blood plasma and 36 urine samples from the sysVASC cohort and 1000 blood and urine samples from the KORA cohort.
These data were analysed using a combination of tools including, cross-omics comparisons, literature mining, pathway analysis, and investigation of relevant existing data-sources, highlighting a series of interesting targets (described in WP4-WP5).
As an example of the comprehensive analysis conducted at the protein level, a snapshot from the comparisons of the results from central versus peripheral vessel analysis is shown below (Table 3).
Table 3: Examples of consistent proteomic differences between CVD cases and controls in central vessels, peripheral vessels and pooled (central plus peripheral vessel) analyses. Based on an extensive literature search all of the shown proteins have already been described in the context of CVD, supporting the validity of the approach. Red colour indicates up-; green colour: down-regulation in cases versus controls. In blue: the difference reached statistical significance in the particular comparison.
Symbol Name central vessels peripheral vessels pooled analysis
Number of peptides ratio Mann Whitney p-value Number of peptides ratio Mann Whitney p-value Number of peptides ratio Mann Whitney p-value
PDLIM7 PDZ and LIM domain protein 7 3 0.42 3.70E-03 3 0.67 5.55E-02 2 0.56 2.33E-03
HSPB6 Heat shock protein beta-6 2 0.77 6.87E-02 2 0.68 3.68E-02 2 0.80 6.21E-02
POSTN Periostin 15 3.96 1.82E-02 18 1.26 3.79E-01 15 1.77 4.30E-02
ATP1A1 Sodium/potassium-transporting ATPase subunit alpha-1 2 1.95 2.69E-02 4 1.61 1.33E-01 2 1.74 7.77E-03
TTN Titin 15 1.38 4.03E-01 15 1.23 5.69E-01 11 1.52 3.49E-02
FBLN1 Fibulin-1 3 1.82 2.29E-03 2 1.72 1.05E-01 2 1.51 5.82E-02
Verification of findings
In view of further verifying the identified proteins analyzed via proteomics, Western Blot (WB) and/or immunohistochemistry (IHC) were performed on A. iliaca communis samples of the sysVASC patient cohort. At first, all A. iliaca communis samples were pathologically evaluated via hematoxylin eosin-stained paraffin sections. Samples were classified in 5 groups: Group A with completely thin intima; Group B with slightly thickened intima; group C with a thickened and partly strong thickened intima; Group D with a thickened intima and calcification in the intima, and Group E with a thickened intima and calcification in the intima and media (Figure 5).
Figure 5: Histopathological classification of hematoxylin and eosin-stained paraffin sections.
Using these well characterised samples several of the proteomic changes were confirmed including the shortlisted targets for interference studies (described in WP5).
1.3.3. WP3: Statistical Modelling and systems Medicine Analysis
Eagle Genomics (EAG): Part of Eagle’s role in the consortium was to combine the data from different analysis produced by the different partners. Some of these analyses enable the identification of potential new targets/genes potentially involved in the understanding of cardio-vascular disease progression. In total 9 new targets were identified by the consortium. At this stage Eagle was able to apply and refine its unique data valuation technology in order to help in the prioritisation of the most promising targets for further validation work. This technology enables the capture of subjective criteria (e.g. disease association, drugability, toxicology profile, novelty) provided by domain expert and represent in a very explicit and objective way the decision-making process. The sysVASC project has enabled Eagle to focus the application of its technology on a very concrete use case, very applicable and valuable to the pharmaceutical industry.
University of Glasgow (GLA): The stroke-prone spontaneously hypertensive (SHRSP) rat and its reference control strain, the Wistar Kyoto rat (WKY), are well-characterised models widely used in the investigation of cardiovascular disease. By comparison of the signal pathways and networks in these models it may be possible to elucidate the mechanisms which lead to hypertension and cardiac disease. We have used tissue proteomics methodologies combined with label-free protein quantitation to investigate the mechanisms underlying cardiovascular disease in SHRSP and WKY rats, with and without salt loading.
By utilising a tissue proteomic approach to both SHRSP and WKY rat aortas we have described a total of 346 significantly differentially regulated proteins. Of these, 40 were downregulated in SHRSP exposed to 1% salt while 70 proteins were upregulated under the same conditions. There were several proteins in this dataset which are known to be involved in cardiovascular disease which increases confidence in the dataset. In the WKY dataset there were 96 proteins downregulated in the presence of 1% salt and 140 proteins upregulated.
We also investigated the interactions between proteins in the datasets. We utilised a publically available database (String DB) in order to investigate these interactions. Our results highlight potentially differentially regulated pathways and signalling networks. By mapping the differentially regulated proteins in this way we are able to draw comparisons between different models of cardiovascular disease.
Helmholtz Zentrum München (HMGU): In partnership with INSERM and Biocrates, HMGU wished to determine metabolites indicative of cardiovascular risk (CVR, here using carotid intima media thickness (cIMT) as a surrogate) and diabetes in humans using the Biocrates metabolomics platform, and to determine which animal models most closely mimic these conditions. Regression analysis was used on the metabolite profiles of the study participants of the Cooperative Health Research in the Region of Augsburg (KORA) F4 population-based study (2006-2008, N=2714 for cIMT, N=3041 for diabetes) to determine differences between individuals at risk for cardiovascular disease and those showing less risk. The profiles were also compared for individuals with and without prevalent diabetes. The results were compared to those for knock-out mouse models, rat models and klotho mouse models. For both human and animals, metabolites were ranked according to their differential concentrations in the case group compared to the control group. The list rankings for each animal model were then compared to the list ranking for humans. A “top” model for CVR risk was determined, while for diabetes no model showed much similarity to humans, likely given the different types of diabetes represented (type 1 in animals, primarily type 2 in humans).
HMGU BGC was responsible for method developments and data integration of non-targeted metabolomics datasets. Non-targeted metabolomics aims at the detection and analysis of all metabolites and small molecules in a living organism so as to enable the modelling and understanding of diseases and perturbations, which is to complement strategies for metabolic control. Despite decades of analytical and technical advances in terms of sensitivity and specificity, data analytical strategies are still in their infancy in comparison with other ‘omics’. In non-targeted metabolomics analysis, typically more than 70% of signals with assigned molecular formula (e.g. C6H12O6 for glucose), but systems biology can only perform once data can be mapped to biological prior knowledge. We developed new general strategies for mass-spectrometry-based metabolomics that incorporate all signals with known molecular formula into the evaluation and interpretation of metabolomic studies. We used mass differences known to occur between metabolic m/z signals as a result of biochemical reactions to build up data-driven metabolic networks. We extend the analysis of such networks by focusing on the occurrence patterns of these known biochemical reactions, to determine how an organism builds up its body from known metabolites. This strategy – called mass difference enrichment analysis – immediately connects the metabolomic data to transcripts, proteins, disease phenotypes and other metadata. Analytical methods developed by BGC use the entirety of metabolomic datasets, therefore improve interpretability of yet unknown biochemical relationships and open a new door for systems biology to make use of metabolomics.
Mosaiques Diagnostics & Therapeutics AG (MOS): Clinical proteomics aims at developing protein/peptide biomarkers that either as stand-alone or in combination with other clinical variables will improve current medical practices. These biomarkers can aid in the management of chronic diseases i.e. diagnosis, and prognosis. However when combined with clinical parameters, high-resolution mass spectrometry proteomics generate large and complex data requiring sophisticated and robust computing statistical analyses.
Within sysVASC, we compared the effectiveness of different statistical methods to enable urinary biomarker discovery. Additionally, we investigated the performance of urinary biomarker signatures also called classifiers established by different modelling approaches for the diagnosis of coronary artery disease (CAD). In cohorts of 197 and 368 participants, respectively assessing the diagnosis of CAD and the prognosis of acute coronary artery syndromes, urine samples were analysed via capillary electrophoresis coupled to mass spectrometry (CE-MS). For biomarker discovery, different statistical methods including the wilcoxon rank sum test, t-score, cat-score, binary discriminant analysis and random forests were applied and compared. Findings depicted differences in biomarker identification depending on the statistical approach used. These differences in biomarker signatures did not translate into significant differences in the performance of diagnostic or prognostic classifiers modelled by support vector machine, diagonal discriminant analysis, linear discriminant analysis, binary discriminant analysis and random forest. Thus, findings showed that above mentioned statistical methods can be successfully used to manage proteomics data.
The analysis of mass spectrometry (MS) data to enable peptide identification with high confidence is very complex and challenging. Currently available solutions do not enable harmonising and clustering the data obtained from MS analysis. However, such harmo-nisation is essential to detect significant changes with sufficient confidence. Therefore, we developed a software solution for handling high-resolution proteomics data to enable peptide identification with high confidence. Altogether, efforts generated within this work package enabled to better understand and manage large proteomics data.
Swiss Institute of Bioinformatics (SIB): The aim of our activity was to use network-based dynamical models to identify and predict key molecular events – and possible targets for therapeutic intervention – in the evolution of cardiovascular pathologies, particularly athero¬sclerotic plaques formation.
We have chosen to adopt a logic-based modeling formalism that avoids the need for explicit determination of kinetic parameters. Our logical model is based on a prior knowledge network (PKN) that combines human and animal data from over 200 papers and includes a total of 729 nodes (including 432 proteins and 297 other entities such as metabolites and biological processes) joined by a total of 3,406 edges (regulatory interactions). Experi¬mental proteomic datasets provided by the other sysVASC partners were used to construct subnetworks from this PKN for animal and human systems; Boolean models generated from these subnetworks were then used in dynamical simulations to identify critical nodes that are potential therapeutic targets. These models generated two groups of attractors characterized by two clusters of nodes with opposing behaviors; one cluster enriched in nodes implicated in lipid metabolism (many of which are known to be negatively correlated with CVD), and a second opposing cluster enriched in nodes implicated in inflammation (many of which were positively correlated with CVD). In silico perturbation experiments were also used to identify nodes that were able to switch the network from a diseased state to a healthy state and vice versa.
University College of Dublin (UCD): UCD provided data analysis and statistical modelling in order to identify biomarkers for the prediction of risk and early diagnosis of cardiovascular disease from proteomics and metabolomics data. Biological data are usually noisy and especially data from patient samples can contain high variability. Therefore, a focus was on identifying the most suitable analysis methods. For this we compared several statistical classifier models on urinary proteomics data. Different approaches produced different biomarker patterns, albeit of similar performance. Thus, different methods can be combined to produce master patterns. Another approach focussed on dealing with missing data points in mass-spectrometry (MS) based proteomics, which is a particular issue with proteins close to the detection limit. We developed a Bayesian algorithm that exploits this knowledge and uses missing data points as a complementary source of information to the observed protein intensities in order to find differentially expressed proteins. The algorithm consistently outperformed other methods. We then applied these methods to analyse MS based proteomics data from urine, blood vessel and heart tissues. The results suggested that aberrations in extracellular matrix (ECM) and complement activation are enriched in diseased blood vessels. This indicates that remodelling of cell adhesion and blood clotting abnormalities are main contributors to the development of cardiovascular disease. In diseased hearts suffering from dilated cardiomyopathy a vastly enhance glycolysis was observed. On the other hand, in ischaemic heart disease oxidative phosphorylation and mitochondrial structure were affected.
University of Manchester (UNIMAN): The University of Manchester has worked closely with INSERM Toulouse on the problem of identifying the proteases that lead to the peptides found in urine and other body fluids. Peptides are parts of proteins formed as proteins are degraded; one mechanism for this degradation are enzymes known as proteases that cleave proteins into peptides. This cleavage is a normal part of the body's processes; however, a perturbation to these processes can result in an atypical peptide profile. Knowing which proteases are implicated in peptide production can give insights into the pathways and processes occurring within the body.
Identifying peptides and the proteins whence they came is a long studied problem. Less studied is the problem of the proteases implicated in peptide production. We have developed a Web based tool called Proteasix that tackles this question. The user provides a list of peptides and the proteins whence they come. The end of each peptide implies a cleavage point for a protease. Protasix re-constructs those cleavage sites and then matches those cleavage sites to known proteases for those sites and proteases predicted to cleave at those sites. The result is a list of proteases linked to the proteins and the peptides produced. The proteasix tool has been used successfully in analysing many data sets.
1.3.4. WP4: Evaluation of similarity of existing CVD model systems to human disease
Objective
The major objective of WP4 was to provide animal models of CVD and perform detailed molecular characterisation using omics analysis of animal tissues and body fluids and evaluate the similarity of human CVD and the selected animal models.
sysVASC’s animal models
A wide range of animal models was generated and analysed including hypertensive rats (SHRSP) and mice (angiotensin II infusion), atherosclerotic mice (Ldlr-/- and apoE-/-) with and without diabetes, an atherosclerotic plaque rupture model and a model of vascular calcification (Klotho mice). Some typical lesions in these models are shown in Figure 6.
Figure 6: Typical CVD lesions in the sysVASC’s animal model suite. A) Thickening of the vessel wall in hypertensive rats. B) Plaques in atherosclerotic mouse models which is exacerbated by diabetes. C) Calcification in the Klotho mouse model (in red). D) Increased medial thickness in mouse hypertensive model (in red) and E) Vessel segment I represents unstable/rupture-prone atherosclerotic plaques characterized by disruption of fibrous caps and luminal thrombosis in the plaque rupture model.
These models were used to produce an impressive amount of omics data that were used to describe the animal model similarity to human CVD and select appropriate CVD models for drug studies (WP5).
Urinary peptides
As a proof-of-concept study we showed that we could use urinary peptides to determine the similarity to human CVD of the atherosclerosis/diabetes mouse models. Comparing the urinary peptidome (ortholog peptides) of these mice to two previously established peptide models of human CVD (CAD238 classifier, Zimmerli et al., Mol Cell Proteomics 2008) and the ASCP75 classifier (Htun et al., PLoS One. 2017) it appears that the diabetic Ldlr-/- strain is most versatile since it might serve as a model of CAD at weeks 15 and as a model of ACSP at weeks 22, while the Apo-/- strain seems only a good model of CAD (Figure 7).
Figure 7: Abundance of ortholog urinary peptides in different atherosclerotic/diabetic strains in time.
STZ, streptozotocin (inducing diabetes).
Vessel proteins
Comparison between the mouse and rat model proteomic data with collected proteomic data from human vessels (WP2) was also performed. Mouse and rat homologues of the human proteins were retrieved manually using databases including Uniprot, the Mouse Genome Informatics database, the Rat Genome Database, and HGNC (HUGO gene nomenclature committee). In total, 109 differentially expressed proteins (between CVD cases/models and respective controls) were found in human and at least one animal model. Interestingly, multiple shortlisted candidates for IHC (described in WP2) including those further targeted via functional studies (such as KDM5D, described in WP5) were also detected changing in multiple animal models. As an example, a snapshot of results from this analysis is shown in the Table 4 below.
Table 4: Example of animal model-human comparisons at the protein level. Differentially expressed proteins in vessels of CVD cases (or models) versus respective controls were compared to identify “conserved’’ (inter-species) molecular changes associated with CVD. “yes” indicates differential expression with agreement in expression trend (up- or down- regulated) with the humans, whereas “no” indicates that differential expression was not observed.
Pathway analysis
Similar comparisons between rodent models and humans at the pathway level but based on proteomic data were also performed. In an assessment of the similarity of the sysVASC animal models with human disease based on vessel proteomes, all molecular identifiers were initially mapped to the CluSO (Clustered sequences and Orthologs) and OMAP (Ortholog mapping) databases, which are multi-species sequence and chemical compound cluster databases, in order to facilitate a comparison between animal (rat and mouse) and human data. This was followed by merging of repeated entries in each individual dataset, and alignment of all data into one discovery matrix. Simultaneously, to allow pathway mapping to existing composite large-scale metabolic and signalling cascade pathway maps, all data was mapped to mouse. All molecules (mouse orthologs) were loaded according to the comparisons in Pathvisio (v3.2.1) regardless of p-values and fold-changes and mapped to reference pathway maps using red (up-regulated) or green (down-regulated) colours (grey denotes molecules that were not present in the datasets). Large-scale maps used were derived from analyses of mouse diabetic aortic vessels, of mouse kidneys from an acute kidney injury model and ischemia reperfusion injury of the rat heart (Figure 8). However, it should be noted that such maps are a collection of reactions implicated in the diseases and do not reflect a specific disease model/state.
Figure 8: Large-scale pathway mapping. Composite map 1 (AKI, left), composite map 2 (diabetic vessel, middle), composite map 3 (Ischaemia reperfusion injury, IRA; right). Input are ALL identified molecules; red: up-regulated, green: down-regulated; grey: molecule not present in discovery matrix).
Plasma metabolites
The metabolomics data obtained with AbsoluteIDQ p180 kit and Bile Acid kit for ApoE and Ldrl KO mice with and without streptozotocine (STZ) treatment were analysed in order to identify mouse models that have closest resemblance to human diseases. Initial data analysis (Figure 9) demonstrated that the STZ treatment drastically altered the metabolic profile starting from week 15. Additionally, the ApoE KO mice with STZ treatment showed similar metabolite profile as Ldlr KO mice without treatment. A literature search was per-formed and further analysis was performed focusing on pathways and pathophysiological process reported to be altered in humans. The alterations in following pathways / patho-physiological processes were observed: hexose levels; lipid metabolism; inflammation; amino acid metabolism; arginine metabolism; acylcarnitine metabolism; beta-oxidation; bile acid metabolism.
Figure 9: Heat map of the mouse model data obtained with AbsoluteIDQ p180 kit and Bile Acid kit. Streptozotocine treatment drastically altered the metabolic profile starting from week 15.
1.3.5. WP5: Interference studies in in vivo CVD disease models
Objective
Prioritise targets, determine effects of in vivo interference on the development of CVD and define the molecular mechanism of action of interfering compounds.
Drug repurposing based on proteome human-based CVD signatures.
We used the prediction tool CMAP (Connectivity Map: http://portals.broadinstitute.org/cmap) to identify a drug potentially active in reversing the human CVD protein profile (data obtained in WP2). This in silico approach allowed us to establish a list of 53 drugs 9 of which were considered of potential clinical interest based on literature and clinician practice. One of the top compounds (for confidentiality issues called “AUFF”) was tested in vitro and in vivo for its effect in the development of CVD.
In the first step we selected the most appropriate preclinical animal model of CVD in which AUFF efficiency should be assessed. For that we compared the proteins targeted in silico by AUFF with those modified in the 11 preclinical sysVASC mouse models (described in WP4). This led to the selection of the ApoE-/- mouse model as the model where AUFF should have the largest impact.
ApoE-/- mice were treated for 10 weeks with AUFF. Interestingly, AUFF treatment was associated with a significant reduction in lipid accumulation in plaques (Figure 10).
Figure 10: AUFF treatment for 10 weeks reduced lipid accumulation in aortic roots of ApoE-/- mice.
Immunostaining analysis showed that aortic roots from AUFF-treated mice exhibited a significant reduction in collagen-III protein expression when compared to vehicle-treated mice (Figure 11). In contrast, no change in αSMA (alpha smooth muscle actin) was observed (not shown). We conclude that the anti-atherogenic activity of AUFF was associated with a reduction of fibrosis independent myofibroblast activation.
Figure 11: Collagen-III (Col-III) staining of aortic roots from ApoE-/- mice after 10 weeks treatment with AUFF (n=7) or its vehicle (n=10). *: P<0.05.
AUFF also showed a most potent effect during calcification in vitro in vascular smooth muscle cells (Figure 12). The effects of AAUFF are now being confirmed in an established mouse model of vascular calcification.
Figure 12: Effect of AUFF on in vitro calcification. A) Alizarin red staining in HAoSMCs following treatment for 11 days with control or with calcification medium (Calc.) without or with additional treatment with AUFF. B) Quantification of calcium content in HAoSMCs following treatment with AUFF. *(p<0.05) statistically significant vs. control treated HAoSMCs; †(p<0.05) statistically significant vs. HAoSMCs treated with Pi (Calc) alone.
We have thus shown that drug repurposing based on protein-based molecular signatures obtained from diseased vascular tissue can identify valid compounds with potential pro-tective effects on CVD.
Novel Targets
Bringing together the high-resolution tandem mass spectrometry analysis (LC-MS/MS) data of thoracic aortas from two atherosclerotic mice models, Ldlr-/- and ApoE-/-, and from the human central and peripheral vessels and literature, KDM5D, a specific protein involved in epigenetic regulation, was highlighted as being upregulated in CVD cases versus controls. KDM5D (also known as JARID1D or SMCY) is one of the four members of the KDM5 family of histone demethylases that specifically demethylate trimethylated and dimethylated Lys-4 of histone H3 (H3K4me3). Preliminary data (Figure 13) suggest a lower abundance of H3K4me3 in vascular tissues from patients with atherosclerosis compared to controls, potentially confirming the observed elevated expression of KDM5D in atherosclerosis. As inter-individual variability may be seen, this preliminary observation will have to be confirmed following the analysis of a higher number of clinical samples.
Figure 13: A) Western blot analysis of Trimethyl-Histone H3 (Lys4) in protein samples from i) peripheral vessels and ii) central vessels from patients with atherosclerosis and organ donors without cardiovascular background. Total Histone H3 was used as loading control. B) The quantification of the proteins from both central and peripheral vessels was performed by using the Quantity One software (BioRad) and the results were normalized to total H3.
Given the observed up-regulation of KDM5D and the downregulation of its substrate H3K4me3 in disease versus control, we sought to evaluate the effect of KDM5D inhibition in HUVEC cells. Inhibition of KDM5 resulted in an increase in the levels of the trimethylated state of lys4 of histone 3 (H3K4me3) as supported by Western Blot analysis (data not shown). Interestingly, the KDM5 inhibition also resulted in a significant reduction, by at least 40% in both the proliferation and migration capabilities, with a further reduction by at least 20% in the number of nodes, junctions or segments developed by the HUVEC cells. As an example, representative images from the investigation of the tube forming ability of the cells are shown below (Figure 14).
Figure 14: Representative images from the study of impact of KDM5D inhibition on the tube-forming ability of HUVEC cells. Three independent experiments were performed in duplicate each time.
Collectively, these data suggest a role for KDM5 histone demethylases in the development of atherosclerosis likely via affecting the H3K4 methylation. Further experiments investigating the impact of the KDM5 inhibitor on plaque formation are required.
Another target that derived from comparison of omics datasets across species and different types of vascular diseases as described is sirtuin-1. Sirtuins are critically involved in the regulation of mitochondrial processes and have previously been implicated in ageing, inflammation and apoptosis – all of which are features of vascular diseases. The widely used antioxidant resveratrol is known to be an unspecific activator of sirtuins. This has prompted us to study the effect of resveratrol on vascular function and structure in the SHRSP, a well established rat model of hypertension that has also informed the proteomic data leading to discovery of differentially regulated sirtuin-1 in our datasets.
Animals were exposed to 10 weeks of resveratrol treatment (130 mg/kg/day) from 5 weeks of age and underwent weekly blood pressure measurement. Wire and pressure myography have been conducted on mesenteric arteries. Our data suggest that the Sirt1 activator resveratrol has no statistically significant impact on blood pressure in the SHRSP. However, on a structural level, resveratrol-treated animals show an improved distensibility profile (stress - strain relationship) when compared to controls, which suggests improved vascular remodelling (i.e. reduced vascular stiffness in the resveratrol treated rats) (Figure 15).
Figure 15: Pressure myography in mesenteric arteries from resveratrol and vehicle treated SHRSP rats. A) external diameter, B) internal diameter, C) lumen diameter, D) cross sectional area, E) wall stress and F) vessel distensibility.
1.4. WP6: IP Management and Exploitation
Coronary artery disease (CAD) is a major public health concern in developed countries. Although a significant improvement in the management of CAD was observed, the prevalence of CAD is still on the rise due to demographic transition and increased survival rates. The aetiological diversity of CAD makes it challenging to fully decipher complex pathological processes based on single biomarkers. It is therefore obvious that innovative methods that will improve the management of CAD are urgently needed. Despite the socio-economic burden caused by CAD, the development of innovative therapeutic compounds has been moderate due to a limited understanding of primary mechanisms associated with the disease.
The scientific aims of sysVASC were to address current obstacles in the field of CAD including:
• Limited understanding of molecular mechanisms associated with CAD.
• Limited access to human samples.
• Lack of suitable mouse models for therapeutic interventions.
• Lack of efficient drugs targeting causative molecular mechanisms of the pathology.
Beyond these aims, sysVASC additionally focused on exploring the translation of research findings into health care applications. To achieve an effective translation and based upon high level of interaction between partners, a work package was dedicated to intellectual property management and exploitation. To translate scientific discoveries into clinical implementation, several objectives were defined at the beginning of the project including:
• Design of activities aiming at the protection of intellectual property rights.
• Direction of identified targets towards effective exploitation by appropriate third parties.
• Setup and update a database of state-of-the art in cardiovascular disease drug development and relevant (pre)clinical trials.
• Efficient exploitation of results during and beyond the lifetime of the project.
During the project, efforts were generated by partners involved in this work package to achieve the above mentioned goals. This led to the exploitation of sysVASC findings during and even beyond the duration of the project. At first, a strategy to optimise exploitation was setup and the main principles for the management of the intellectual property rights were as follow:
• Results were openly disseminated in parallel with appropriate protection of intellectual property to encourage the promotion of scientific results during the project. In case of publication, reasonable notification and deferral procedures to ensure intellectual property protection were set in place and followed.
• Background intellectual property remained the property of the partner (s) that brought it to the project. Access rights to Background intellectual property was granted on the basis of: 1) project use (‘for implementation’) and 2) exploitation (‘for use’) to another partner needing to exploit the data in accordance to agreement between concerned partners. As an example of Background intellectual property, Mosaiques Diagnostics GmbH had access to the intellectual property rights on biomarkers for cardiovascular disease prior to the project.
• Results of sysVASC (Foreground intellectual property) were owned by the partner(s) responsible for generating the findings of interest.
Generally, issues relating to intellectual property rights were managed in such a way that exploitation decisions were made to serve the needs and goals of the partners.
During sysVASC, novel therapeutic interventions to tackle CAD were identified. As a result, potential targets for intervention in CAD were identified. To optimise the exploitation of the findings, a patent search was performed with the compounds of interest to further investigate the novelty of the results. As translation of results in industrial application was envisaged, this was a very curial point that was carefully explored in the exploitation strategy. The patent search that was made, covered biomarkers and therapeutic targets described in the context of cardiovascular diseases. Specialised websites were used to screen for European as well as international patents that were either granted and/or published. Additionally, patent search and frequent updates enabled to ensure the potential financial value of identified therapeutic targets. Results of the patent search were helpful to redefine when necessary a relevant context-of-use of a targeted molecule or compound.
The aim of sysVASC was to exploit this unique know-how and transform it either into highly innovative products or services to generate business opportunities. The unique sysVASC know-how or approach (Figure 16) offers solutions to assist in a more successful way drug development which is of high relevance to pharmaceutical industries.
Figure 16: The sysVASC approach. Datasets and related clinical data were combined with existing knowledge on CAD from literature and omics data repositories. Data were analysed using appropriate bioinformatics and statistics tools to generate an initial model of relevant molecular and pathway changes in CAD. Key components of these pathways were verified in appropriate human samples. In parallel, animal models were investigated for molecular homology, enabling to choose the model system best reflecting the human disease. Intervention studies in this model enabled assessing its relevance to disease, guiding the development of drugs directly addressing molecular causes of CAD.
As result, pharmaceutical companies represented the targeted market group and were considered as customers of choice. Benefits of the know-how for pharmaceutical companies included:
• Access to unique human vessel tissue samples to identify adjacent proteomic changes relevant in CAD.
• Access to the unique surgical tandem stenosis mouse model reflecting all stages of human atherosclerosis.
• Access to a unique database of urine samples to help achieve a systemic assessment of CAD-associated changes and their validation.
• Access to comprehensive proteomic analysis using cutting-edge technologies for a better and an in-depth protein characterisation.
• Access to proprietary software solutions to evaluate proteomic data.
• Access to proprietary tools to facilitate the selection of animal models best displaying the human disease-phenotype based on proteomics.
• Access to proteomics and bioinformatics tools to facilitate drug target identification.
• Access to a unique proteomic expertise breaking through the norms of protein identification and characterisation by further exploiting the information to identify potential drug targets and select appropriate animal models for intervention studies.
Furthermore, to achieve effective dissemination in the above mentioned market within the duration of the project, additional efforts led to the development of a marketing plan.
To generate business opportunities with pharmaceutical companies, a dissemination database was developed and updated. The database included collected information on current state-of-the-art in cardiovascular disease drug development and relevant clinical trials. Additionally, other relevant information including the name of the principal investigator of a trial, and/or the name of an employee working in the area of cardiovascular disease in a pharmaceutical company along with a relevant email were included. Using the email addresses included in the database, people of interest were contacted to initiate business opportunities. The dissemination database was a key tool to promote exploitation of results because it helped reach the targeted audience.
To further contribute to the successful exploitation of the project, a proposal collecting key findings and highlighting the unique approach used to improve drug development was generated at the end of project. The proposal summarising results generated within sysVASC was used as a basis for contacting pharmaceutical industries and establishing new collaboration to further validate novel therapeutic targets in clinical studies. Using contact details collected in the dissemination database, the proposal was disseminated to people of interest. As this process is ongoing, potential business opportunities emanating from this initiative can only be evaluated after the project. However in the meantime, a negotiation strategy was developed to ease conversation with pharmaceutical companies. Briefly, the negotiation strategy will include preparation, discussion, proposing and bargaining phases that will help secure a successful collaboration.
Altogether, the sysVASC project generated exploitable intellectual property proposing novel strategies (i.e. therapies) to improve the management of CAD. Exploitation of these resources enabled to initiate business opportunities.
Potential Impact:
1.5. Potential Impact
Within sysVASC, we implemented a systems biology approach to gain knowledge on the molecular pathways underlying CVD. The multidisciplinary effort benefited from (i) excellent preclinical and clinical research centres, (ii) cutting edge omics technological platforms and (iii) unique bioinformatics and statistics expertise. With the broad, essentially hypothesis-free approach, sysVASC was designed to become a showcase for the application of systems biology in clinical research and development. The integration of previously disparate data sources generated a unique comprehensive expandable resource for identifying and validating novel key molecular targets for treatment, thus tackling the lack of interventions specific for CVD and having major scientific, societal and economic impact. Identified targets for CVD treatment were advanced towards translation into novel therapeutic approaches and may even result in additional products such as biomarkers for diagnostic/prognostic purposes, thereby relieving a major burden to patients and health care systems and promoting CVD treatment towards personalised patient management and individualised targeted therapy.
1.4.1. Potential impact of the results of WP1
WP1 has developed a number of tools that facilitate the integration of multi-omics data across different datasets in clinical cohorts. The tools that have been developed and described above are available to the scientific community. Two resources should be mentioned in particular in this context.
First, the CVDKB is accessible to researchers worldwide (http://cvdkb.org/) and supports data analysis and integration. The experiments span over different biological levels (genes or proteins), different techniques, different species (mouse or human), and different sample types (blood or coronary artery). Experimental data derive from published articles in Pubmed, all related to coronary artery disease. The data have been extracted manually from the figures or from the supplementary data when possible. Microarray data available in the gene expression omnibus (GEO) were reanalysed in house using standard R scripts (GEO2R). Altogether, 34 publications met the inclusion criteria and 26 were excluded.
Second, e[ADS] (developed by Eagle Genomics) has several impacts for clients in academia and industry:
• Improved time and ability to insight through improved productivity of researchers.
• Faster identification and generation of successful product innovation and insight, leading to reduced time to market.
• Competitive advantage through cutting edge research, driving innovation in the product pipeline.
• Efficient compliant, documented, evidenced, and enforced data sharing process; effective (time & cost) collaboration between vendors and internal teams.
• Systematic data sharing and re-use, driving faster unique cross-functional insights and reducing costs.
• Rapid seamless integration of diverse data sets for scientific analysis.
1.4.2. Potential impact of the results of WP2
• Generation of an invaluable resource of well characterized clinical samples and comprehensive molecular data to support further integrative studies on CVD in the future.
• Lists of verified proteomic changes in human CVD of high biological interest, whose further investigation as potential therapeutic targets is warranted.
• Optimized, reproducible procedures for sample preparation and workflows for proteomics data analysis of starting material of limited amount, disseminated to the community to increase comprehensiveness and data reproducibility of similar pre- and clinical proteomics studies in the future.
1.4.3. Potential impact of the results of WP3
Eagle show cases this data valuation use case at any customer opportunity to demonstrate its strategic potential in target portfolio management and has already opened up other opportunity where the prioritisation of other “entities of interest”, e.g. best patients for stratification, best indications for drug repurposing or best natural ingredients for new skin care cream formulation.
HMGU: Metabolomics represents the closest omics level to clinical phenotype and can inform on both products of disease or possible drug targets. The qualitative and quantitative knowledge of the metabolic similarities between animal models and human cardiovascular risk can allow us to better plan future cardiovascular disease studies using the most appropriate model, either targeting a small collection of specific metabolites or utilizing the metabolomics profile as a whole.
MOS: Proteins regulate all biological functions and allow an optimal representation of physiological, environmental and pathological variations occurring over time. Thus, proteome analysis is a suitable approach providing necessary information on molecular pathological processes and enabling a better disease characterisation. Recent advances in proteomics and the development of robust tools (mass spectrometry (MS)-based) for an in-depth characterisation of proteins now enable comprehensive investigation of proteomic alterations in complex biological samples. However, identification of peptides with high confidence remains a challenge. In this work package, we have developed a software solution to improve peptide identification. Therefore, accurate protein identification will contribute to a deeper understanding of molecular mechanisms associated with diseases and improve disease management. As a result, the impact of the software solution will move beyond the scope of sysVASC.
SIB: Our computational simulations have allowed us to identify a list of potential therapeutic targets that can be validated experimentally. A promising target that was identified in all analyses using models built using proteomic data from animal and human tissues was the protein SIRT1, which has previously been shown to have a protective effect on cardiovascular disease and which in our simulations appears as a promising therapeutic target with the ability to reverse the disease phenotype when overexpressed/activated.
UCD: These studies have provided new analysis methods for proteomics data from human body fluids and tissues. The results suggest new biomarkers for the diagnosis of cardiovascular disease and different forms of cardiomyopathies.
1.4.4. Potential impact of the results of WP4
Comparison of human disease and preclinical animal models of CVD at the molecular level will lead to the:
• definition at the molecular level of the most appropriate models representing certain aspects of human CVD for subsequent molecular and interference studies;
• development of “humanized” readouts in mice that are supposed to significantly speed-up the translation from the preclinical to the clinical phase since we anticipate that successful drugs as monitored based on these humanized readouts will remain successful candidates during the move from animals to humans;
• towards reduced use of animals models in line with the 3Rs European policy
o For drug testing, the in silico selection of the model most closely mimicking certain aspects of human disease among a panel of available models of CVD avoids testing blindly all available CVD models (Replacement and Reduction).
o In addition, the development of non- or minimally-invasive (humanized) readouts will lead to the definition of alternative end-points (Replacement) and will allow for longitudinal follow-up during drug-testing without animal sacrifice (Reduction).
1.4.5. Potential impact of the results of WP5
The potential impact of WP5 is threefold:
• Conceptual: show-casing that the pipeline as developed in sysVASC is feasible. The combination of the sysVASC human omics data, bioinformatics pipeline and the sysVASC’s animal CVD model suite has allowed the identification of new drugs and targets. Such pipeline is applicable to any disease where omics data can be obtained.
• Identification of novel CVD targets: this pipeline focusing on the molecular mechanism of CVD has generated a number of original and very diverse targets with a focus on calcification, ECM and epigenetic modifications. This shows that detailed insight in the molecular mechanism of disease obtained by omics analysis allows the identification of very diverse novel targets of disease.
• Drug repurposing speeding up in vivo validation: in silico analysis of reversal of CVD signatures by a large panel of drug signatures generates compounds that can be readily tested in vivo.
1.4.6. Potential impact of the results of WP6
The sysVASC project aimed at providing innovative approaches to facilitate drug discovery by identifying novel therapeutic targets. The novelty of the proposed approach relied on the fact that “omics” platforms combined with literature mining were used to identify drug targets after application of robust bioinformatics tools.
Several pharmaceutical industries are actively involved in the development of novel drugs to improve the management of disease. These industries include Novartis, Pfizer, Roche, Sanofi, Merck & Co., and many others all wanting to develop the most effective drug at a low cost. The sysVASC approach makes it possible for pharmaceutical industries to achieve these goals. This can be achieved using the sysVASC approach by identifying CAD-specific biomarker signatures in tissue, blood and urine samples. Together with a thorough investigation of the literature and aided by robust bioinformatics tools, relevant molecular pathways can be investigated leading to the identification of therapeutic targets. The unique biomarker signatures can in parallel also enable the non-invasive selection of animal models and drug monitoring and a reliable patient stratification. Contacting pharmaceutical companies has the potential impact to generate business opportunities and influence the field of drug discovery in cardiovascular disease at large.
Since the sysVASC approach regrouped a multitude of technologies, this enabled a rapid identification of drug targets. Moreover, with the access to animal models, identified targets could be validated. Hence, the approach was time-efficient and cost-effective.
The potential economic benefits for pharmaceutical industries in using the sysVASC approach will thus include:
• The development of more effective drugs. Due to a comprehensive “omics” analysis provided by the sysVASC approach unique disease signatures can be generated, novel drug therapies specific for the treatment of CAD can thus be identified. This will significantly reduce the costs as compounds to be tested in a clinical trial will be of high confidence. For instance, developing a compound with high confidence would have prevented some financial losses already made by pharmaceutical companies in an attempt to improve the management of cardiovascular diseases.
• Better selection of animal models and non-invasive monitoring. The sysVASC approach has the unique advantage to enable the selection of suitable animal models best mimicking the human disease. Furthermore, the efficacy of a therapy can also be assessed non-invasively in the selected animals. This will contribute to the increase rate of success of preclinical studies ultimately leading to clinical studies with high confidence.
Overall the sysVASC approach has the potential impact of generating business opportunities with the pharmaceutical industry. As the sysVASC approach proposes new strategies to improve the management of CAD, it can additionally help reduce the current economic burden caused by CAD.
1.6. Main dissemination activities and exploitation of results
Professional project management and dissemination of results significantly contributed to the successful collaboration of the sysVASC beneficiaries. The Coordinator encouraged all participants to actively disseminate their project aims and results to colleagues, stakeholders and the wider public.
At the beginning of the project, a logo and a corporate identity for the project was created in order to identify the project. It is used together with the Grant Agreement number HEALTH.2013.2.4.2-1 603288 and the FP7 logo (until use of the FP7 logo was officially discontinued) on all printed and electronic material for the public and for any other official contact.
On the web, public awareness of sysVASC was addressed by the project website www.sysvasc.eu. The website content was regularly updated and includes key background information about the project and its beneficiaries. All content was available in two languages: English and German. Additionally, project fact sheets are available in 4 languages for external stakeholders allowing information to be obtained at a glance. The news and media section provided visitors with the latest publications, press releases and dissemination activities of sysVASC, as well as the sysVASC postcard quiz, which was released at the MUG 10-year anniversary, an event reaching 3000 people, to generate greater exposure for the project and increased impact.
The centrepiece of the last website update was the sysVASC 3D video, which clearly presented the background to the project and highlighted some of its successful outcomes.
The Consortium also took measures to provide sysVASC exposure to the public through conventional PR media work and new media, which complemented the public access website, particularly for those that do not have internet access.
The scientific community has a vested interest in keeping up with sysVASC progress and results to build on the body of knowledge that currently exists. This was achieved in part through conventional scientific dissemination means, i.e. peer reviewed publications and presentations at international conferences. 40 peer-reviewed publications were published during the project funding period as part of the project and partners have presented sysVASC work world-wide 20 times at international scientific conferences. These publications also included a significant number with contributions from multiple partners, emphasising the impressive level of cross-consortium cooperation, which the sysVASC project has demonstrated. A number of additional publications are at the preparation or review stage and are expected to be published in the near future.
Two dissemination highlights that occurred during the first half of the project were the Medical university Graz 10th anniversary and ‘Destination LABO’. As part of its 10-year anniversary celebrations the MUG introduced the sysVASC project to the public. 3000 visitors attended the event, held on the 17th October 2014 in the main square of Graz. At the ‘Destination LABO’ event the INSERM’s renal fibrosis laboratory in Toulouse opened their doors to students from primary and secondary schools to present current work regarding chronic kidney disease and its association with cardiovascular disease. A special emphasis was placed on the sysVASC project during the event.
List of Websites:
Name of the scientific representative of the project's coordinator and organisation:
Prof. Dr. Burkert Pieske
Charité – Universitätsmedizin Berlin
Augustenburger Platz 1
13353 Berlin | Germany
Phone: +49 30 450 553702
Fax: +49 30 450 7553 702
E-mail: burkert.pieske@charite.de
Project website address: http://www.sysvasc.eu/
