Final Report Summary - CARTARDIS (Identification and validation of novel pharmaceutical drug targets for cardiovascular disease)
Executive Summary:
The EC FP7 funded project CarTarDis (Cardiovascular Target Discovery) aimed to apply the high quality expertise and resources within its multi-partner consortium to discover novel therapeutic drug targets with a high level of clinical correlation to cardiovascular disease (CVD). The project adopted a strong pharmaceutical work flow in which the 13 consortium partners formed a virtual pharma company in which human cardiovascular cohorts, preclinical models, clinical biobanks and innovative molecular staining methods played a key role. The overall objectives of the project were three-fold: 1. Identify and validate novel CVD drug targets, 2. Develop in vitro and in vivo tools to support CVD drug development, 3. Increase the knowledge and competitive position of the participating partners (notably the SME’s) in drug discovery.
Directed by the experienced pharmaceutical scientists in our team and advised by experienced drug developers and entrepreneurs in our External Advisory Board, we designed a rational approach to accurately reflect the pharmaceutical process, adopting best practices from pharmaceutical industry from its start. First, available and emerging molecular data from three large independent cardiovascular cohorts was mined for statistical association with cardiovascular disease phenotypes, with a preference for genetic associations. A database and target selection workflow was designed that allowed us to select and prioritise candidate CVD drug targets. The selection filters were inspired by the 5R strategy of AstraZeneca and included Genetic/omic evidence, Druggability, Novelty, Feasibility and Pathways. For each preselected target, a target champion was appointed who led a multidisciplinary team of consortium scientists to define the key experiments to generate data to support the validation of that particular target. All teams were guided and supported by the experienced pharmaceutical drug hunter scientists and clinical cardiovascular experts in our consortium, together forming strong teams to reach the objective of identifying novel validated cardiovascular drug targets. Validation of the candidate targets was done by two parallel approaches. Mechanistic validation was performed using novel in vitro (cellular) and in vivo (rodent) preclinical models in which the candidate targets were studied regarding mechanism of action and role in particular aspects of the cardiovascular processes. Clinical validation was based on two unique human cardiovascular tissue biobanks and several high resolution molecular analysis methods using which the spatial expression of the target and target-related pathways in cardiovascular tissue samples was studied on RNA, protein and small molecule level. Particularly the combination and high resolution of the applied technologies was unique. The SME’s in the CarTarDis consortium participated strongly in these two validation activities by contributing their unique technological expertise as part of the validation teams.
The CarTarDis project was successful in several aspects but not in all, in line with the state of the art in pharmaceutical industry. More specifically, we designed a unique interactive database that allowed us not only to summarize key features of preselected candidate drug targets but also to easily adapt the selection criteria to yield a mixed portfolio of drug targets varying in novelty, druggability, omics-based evidence and scientific value. Of the 381 preselected candidate targets, we prioritised 7 targets and formed 7 multidisciplinary target validation teams involving all partners of the CarTarDis consortium. We designed novel in vitro (cellular) and in vivo (rodent) models and multiple novel molecular staining methods that supported the validation studies but also provided added value to the expertise of the consortium partners and to the biomedical CVD field at large. In line with current practice in pharmaceutical drug discovery, we experienced the complexity of human biology and the new data supported further validation for only 3 of the selected targets. For some, we identified downstream proteins as potentially better drug targets than the originally preselected target and focussed our validation experiments on those. We finished the validation projects with clear view on additional experiments that are needed beyond the project to provide ultimate support for the chose targets. Overall, the educational aspect of this process was highly valuable for the participating partners with still limited experience in the pharmaceutical process, one of the key objectives of the CarTarDis project.
Project Context and Objectives:
The development of a novel drug is an extremely challenging task. On average and across various therapeutic areas, the attrition during clinical development is 90%, meaning that only one in ten projects being pursued through clinical testing in patients is successful. The major cause for this high attrition is the proof-of-concept phase, during which the attrition rate is as high as 80% because of a lack of efficacy and/or unacceptable safety liabilities. This leads one to surmise that preclinical studies in pharmaceutical research are insufficient to predict drug action in patients and that high clinical validation of the drug target is key to drug development success. As for other complex chronic human diseases, cardiovascular disease (CVD) is a difficult area to develop pharmaceutical drugs. CVD affect the heart and surrounding blood vessels and can take many forms, such as high blood pressure, coronary artery disease, heart disease and stroke. CVD is a growing medical problem globally with a predicted 23.6 million deaths from CVD in 2030, increased from a 17.3 million death from CVD in 2008 which represented 30% of all global deaths at that time. In the EU, CVD are the largest cause of death in the EU and account for approximately 40% of deaths or 2 million deaths per year. The slow development of the disease (which becomes overt in general above 50 years of age in man and above 60 years of ages in women but may start already at age of 20 ) and the limited availability of clinical samples and translational models has hampered the identification of drug targets and development of efficient therapies. In many cases, cardiovascular association studies have focussed on identifying underlying mechanism of disease and/or clinically relevant biomarkers but not on identification of novel drug targets, and consequently a large part of the obtained data is unsuitable for that purpose. Moreover, clinical material is lacking either in quality or in number of patients, and preclinical models to mechanistically investigate candidate targets are insufficiently translational to the clinic. To increase the chance for successful clinical development of novel therapies in CVD, further innovation is required by combining: 1. Independent and large-scale population studies from which novel targets with strong correlation to clinical phenotypes are deduced, 2. High quality human biobanks to confirm molecular relevance of targets in diseased cardiovascular tissues, 3. Translational disease-mimicking models (cellular and animal) to validate novel drug targets, 4. A stringent pharmaceutical drug discovery process for prioritization of druggable targets. Particularly the druggability assessment of a candidate target, e.g. whether a target protein contains a binding cleft for a drug (be it a small molecule, peptide or protein), is an imperative step in the pharmaceutical process which is rarely being done in clinical association studies.
The CarTarDis consortium brought together the required components mentioned above by careful selection of consortium partners that have a distinct expertise in cardiovascular pharmaceutical drug development and in preclinical and clinical cardiovascular research, in combination with unique resources such as cardiovascular cohort study data, clinical biobanks and molecular analysis methodologies. The required pharmaceutical focus at each step of the project was implemented using the strong pharmaceutical drug discovery experience in CVD among scientists working at the consortium partners. An important concept, in line with the changed pharmaceutical business model, is that these steps can best be executed by a functional network of relevant industry, clinical academic centres and specialized SMEs that brings together the right components and expertise to execute such modular target discovery process. The CarTarDis consortium operationalized a consortium of high quality academic institutions and leading SMEs with key expertise in cardiovascular disease and relevant technologies. It was composed of these parties that each have unique assets and expertise that only in combination can complete the objectives as defined in the project, and pursue the further development and exploitation of its output.
The CarTarDis project objectives were:
• To identify and select candidate CVD drug targets with a strong relevance to clinical phenotypes based on molecular and omics data from three unique large-scale population cohorts.
• To further characterize available, and develop new in vitro and in vivo mechanistic and efficacy models that enable the monitoring of the biological function of CVD targets.
• To develop and optimize cutting edge molecular assays to accurately monitor proteins, RNA and metabolites in preclinical and clinical biosamples relevant to CVD.
• To apply molecular assays to further characterize biosamples of two comprehensive well-annotated human CVD biobanks to investigate and validate the contribution of specific pathways to CVD.
• To apply the in vitro and in vivo models and human biobanks to prioritise and select CVD targets.
• To develop and exploit this newly formed pan-European modular pharmaceutical target discovery network in cardiovascular disease.
Project Results:
The CarTarDis project started 1 October 2013 and ended 30 September 2017. The consortium has disseminated the results and outcomes through presentations, scientific and popular articles and posters. CarTarDis built the foundations for successful development and positioning of emerging drug development tools and molecular methods into cardiovascular research. It also has strongly increased the knowledge of the participating partners in the process of pharmaceutical research and development, providing them a better competitive position in the translational biomedical science field.
• To design a robust target identification and prioritization workflow and criteria based on best practice in pharmaceutical drug development.
The consortium generated a desired workflow with application of filters and criteria based on the most relevant information available of the candidate gene targets identified and linked them to atherosclerosis and cardiovascular events. The workflow included the use of filters that should allow selection of targets for validation in humans to increase the confidence and reduce the risk associated with target selection. Because the biochemical characteristics of individual targets and their relation to different stages of disease development may differ, each candidate target will need to be individually evaluated according to described criteria and specific target-nature characteristics. This assessment includes evaluation of the target causal relation to atherosclerotic cardiovascular disease (ACVD) and if its modulation could be feasible by either small molecular weight compounds, peptides, antisense oligonucleotide or biologicals, such as a proteins or antibodies in a clinical setting (i.e.druggability). We adopted the 5R strategy designed and used by AstraZeneca, as this strategy has shown to be valuable in improving the success of a drug development pipeline. The 5R strategy outlines the assessment of drugs at various levels: Right Target, Right Tissue, Right Safety, Right Patients, Right Commercial Potential. Subsequently, we defined various levels of assessment of our portfolio of candidate CVD drug targets: Druggability, Novelty, Genetic/omic evidence, Feasibility, Pathways. The AZ Pharmconnect database was used to more accurately assess the “druggability” potential of each target. Druggability is a term used in drug discovery to describe a biological target (such as a protein) that is known to or is predicted to bind with high affinity to a drug. Furthermore by definition, the binding of the drug to a druggable target must alter the function of the target with therapeutic benefit to the patient. The concept of druggability is often restricted to low molecular weight organic substances but it has been extended to include large chemical molecules (anti-sense oligonucleotides) and biological molecules such as monoclonal antibodies. Sources of information for this approach included databases of curated drug targets, target screening databases (ChEMBL, BindingDB), literature (PubChem) and manually compiled sets of 3D structure known by the developers to be druggable.
Crucial to the selection of candidate drug targets from the wealth of information collected in the interactive CarTarDis database is the weighing of the selection criteria. Initially we regarded assayability and novelty as the major criteria for target selection, based on the knowledge at that time. However, recent years have seen a dramatic increase in therapeutic opportunities using novel therapeutic agents as RNA interference, small high affinity biological agents and so forth, making the assayability criterium less important. Hence, we adapted the weighing of the selection criteria such to yield a variety of targets with different characteristics and potential.
• To setup a database for all targets and make it available to all the partners.
We have set up a database server for storing of potential CVD drug targets and their annotation data. The system consists of a Unix based server, an SQLite database, and a web frontend based on Python (Django) and Perl. The SQLite database consists of four logical units: first, the target table, which represents the centre of the database; second, several tables to store relevant information on the targets, subdivided into categories, such as genetic information and animal model based information, amongst others; third, the user/institute based part, which allows tracking of the entries in the database; and fourth, a commenting and scoring system to enable a proper prioritization and evaluation of each target, between different partners within the consortium. The accessible web interface for storing and ranking of the potential drug targets was upgraded to adapt the criterion of data protection. Confidential information and the ranking of the targets was access restricted with a password (using the Apache-built-in ‘basic authentification’) and by a twofold IP check (once Apache-based, once interface-based). Additionally, an access-restricted website (https://sites.google.com/site/cartardistargetvalidation/) was set up as knowledge management platform to capture detailed insight into the pre-selected and prioritized targets. Each of the highly emerging targets has been intensively discussed, with ‘pros’ and ‘cons’, underlying mechanisms, druggability potential and other features. Target validation plans, activities and data summaries are shared through this website. This website is frequently updated with the latest available information, with input from multiple consortium members.
We initially collected 381 potential drug targets that were identified either in the public domain (CVD and lipid related GWAS data), or in studies (tissue-based work, animal models, genetic studies) internal to member sites within this consortium. Most of the targets were selected based on genetic evidence (188 associated with CVD and 157 with lipid levels). In general, the targets can be assigned to three groups. (1) high confidence targets at genomic sites linked to CVD in many cohort studies, (2) medium confidence targets showing suggestive evidence for association to CVD, most often in a single study, and finally (3) targets still to be evaluated for genetic links to CVD, but nominated for other reasons including prior art work on human tissues and animal models. The database integrated data from various databases and other sources to provide a comprehensive insight into the potential usability of the collected genes as drug targets. Considering the suggestions of the partners and following the decisions made at the consortium meeting in Gothenburg, we adapted the database to the specific needs of the prioritization for CVD drug targets. This mainly consisted of the commonly agreed relevant data that should be integrated and how it should be represented in the database to provide an informed overview of the targets. Examples of this included the integration of the expert scoring and weighing of the various criteria for drug target assessment. We provided systems level information for each target in order to enhance confidence in the target, provided insights into direction of the effect, uncover disease mechanism(s) and highlight pathways of interest. For this we have carried out an integrative target-specific approach based on various external data sources (e.g. online databases such as KEGG, the Gene Ontology, Ingenuity, Mouse Genome Informatics, the HMGU GWAS database, USCS, Orpha.Net OMIM, patent databases, DrugBank, ChEMBL, Proteinatlas amongst others) as well as internal data sources of human tissue work (molecular networks, regulatory variants like eSNPs/eQTLs, gene expression clinical trait correlations) and genetic data (exome- and genome-wide SNV/SNP data) analysed in a deeply phenotyped cohort (e.g. 13,000 with extensive plaque data). These types of analyses were carried out in both an automated fashion and manual curation for proper data mining. To enable the exploration of the targets with respect to only one of the predefined categories or a customized collection of them, we additionally implemented a filter to only display the categories of interest and fade out the others. This feature is especially helpful in terms of identifying knockout criteria and targets exclusion. Subsequent screenings of emerging literature and updated databases yielded more candidate drug targets that were included in the databases and compared with earlier selections to assess whether prioritisation should be changed. For example, we added 181 genes, which are located in the flanking area 200 kb up- and downstream of 57 SNPs with genome wide significant association to CAD, 511 genes from the CADgene database, 197 genes with blood eQTL values. Moreover, further 756 genes were subjected to our analysis pipeline of which 637 genes were previously suggested to be potentially involved in the development of CVD and 119 genes associated with Alzheimer’s disease that acted as a validation set. The results of our analysis showed that particularly genes from the CAD gene database (average score = 4.55) and those with an eQTL association (average score = 5.00) scored clearly better than those associated with Alzheimer’s disease (average score = 3.89) which is a clear indication that our approach is specific for CVD.
• To identify and select candidate CVD drug targets with a strong relevance to clinical phenotypes based on molecular and omics data from three unique large-scale population cohorts.
We initially pre-selected 381 highly informed potential drug targets from public and proprietary data sources and linked various levels of information to each, including those related to biological function and relation to CVD. We grouped the gathered information into the aforementioned five different categories of relevance. Then, we subjected the information of the candidate drug targets to two approaches of evaluation: an expert-based scoring and an automatic/systematic scoring. Expert scoring was done by experts in the CarTarDis consortium whose appreciation was quantitated and included in the CarTarDis database. However, given the continuous increase in published candidate CVD drug targets and the highly time consuming nature of expert scoring, we aimed to develop a systematic scoring system that would provide an automatic scoring, taking all aspects into account with predefined criteria, weighing of parameters and easy visualisation, that would allow us to test the contribution of criteria filters in an easy yet comprehensive manner. Hence, in parallel with the expert scoring of the first 381 targets, we developed, validated and improved a systematic scoring system. The latter was applied to newly entered candidate targets in the CarTarDis database to assess whether we had been selecting the targets with highest potential.
Expert teams were formed to define evaluation criteria for each of the six defined categories, namely OMICs evidence, novelty, druggability, involvement in lipid metabolism, endothelial dysfunction and inflammation. The expert teams were composed of experts in each domain within our consortium, and asked to apply these criteria to each of the categories and to manually assign scores to every target. The scaling ranged hereby from 0 (no confidence target) to 6 (high confidence) for OMICs evidence and from 0 (no confidence target) to 3 (high confidence) for all other categories. The double weighting of the OMICs evidence was chosen to emphasize the aspired association with the endpoint. In parallel, we attempted to define simple and significant rules that can be applied to easily minable and publically accessible data in order to systematically reproduce the values of such an expert scoring.
With respect to novelty, the experts investigated for each gene if there are patents, in particular with a CVD background, they took a look at the history of investigation a gene underwent, and if it is subject to current research. Our approach to implement this into a systematic score was to mine the PubMed database and extract the number of publications for each gene and specific key-terms. These five key terms were “atherosclerosis”, “coronary”, “cardiovascular”, “hypertension”, and “infarction”. Each of the terms was combined with the gene name and the term ‘gene’ and five scores (one for each term) were calculated. Genes with less than two publications available were scored with 3, genes with less than ten publications were scored with 2 and genes with less than 100 publications were scored with 1. In addition to these five scores, the publication count without the key-term was calculated. Hereby genes with less than 50 publications were scored 3, those with less than 150 were scored with 2 and those with less than 1000 publications were scored with 1. The final automatic score was finally calculated as average of the six previously derived scores.
Concerning the three relevant pathways (endothelial dysfunction, lipid metabolism, and inflammation) the expert scores were mainly based on online literature and databases. In order to automatize this process, we defined specific key terms for each of the three categories and checked their presence in the mined information from PantherDB and Kegg. Detailed information about the used terms can be seen in Table 1.
The experts for the genetic scoring considered several aspects: The presence of a genome-wide significant SNP for CAD, if any genetic variant of this gene is linked to a CVD-related phenotype, a connection to CAC or plaque, and transcriptomic observations. Our systematic approach to implement such a scoring was mainly based on the presence of a known CAD, CAC, MI or CHD associated SNP 200kB up- or downstream. If the significance of the SNP was genome wide (P<5E-08), the assigned score was 3, if the significance was P<1E-05 the given score was 2 and associations with P<1E-04 were given a score of 1. Additionally, an assignment in literature to a SNP with a genome wide significant association to blood lipid levels increased the score by 2 and a genome wide significant blood eQTL value with any of the CAD associated SNPs increased the score by 1 (Table 1).
Finally, the expert score for the druggability integrated complex considerations that were already subject to the scores for the other five categories, as well as tissue distribution of a specific gene, involvement of coagulation, lipolysis, amongst others. Consequently, the implementation of the systematic approach to adequately reflect these aspects was the most challenging, compared to the other systematic scores. To address the link to the endpoint, we used the information on known diseases associated with each of the 321 genes and increased the scores by 1 if any of the diseases was CVD. Furthermore, if a disease with an involvement of blood lipid levels was found, the score was also increased by 1. With respect to the classical definition of druggability- focussing on the question if the gene can be addressed with a small molecule drug – we evaluated the information on the protein type and protein location and increased the scores by 2 if the transcribed protein was located in the extracellular space or by 1 if it was located in the plasma membrane. Additionally, if the genes encoded G-protein coupled receptors, transmembrane receptors, growth factors, transporters, ligand-dependent nuclear receptors, kinases, cytokines, enzymes, and ion channel their score was increased by 1.
The defined process of selection and prioritization, which was a combination of automatic scoring and expert reviews by extensive target background analyses through telephone and live interactions, using the CarTarDis target database as the central shared knowledge platform, was further refined. Correlations between the experts and between expert and automatic scoring were evaluated, in order to make the selected workflow robust and applicable in future drug target selection projects. Overall we obtained good to excellent correlations between the systematic and the expert score. Pearson’s correlation coefficient was above 0.5 for all six categories. The best compliance could be observed for the category evaluating an involvement in lipid metabolism (r=0.81) but also for druggability and novelty, the observed correlations between expert and systematic scores were above 0.65. When comparing the expert scores for different categories among each other, OMICs was least correlated to the other categories. Absolute Pearson’s correlation coefficient was not exceeding 0.2 in the comparison with any other category (Supplementary Table 2). Furthermore, novelty was negatively correlated with all other categories. This mainly reflects that a high degree of novelty consequences a lack of information about a specific gene, as it is not yet fully explored and potential affirmative information is not yet available. In particular, druggability and novelty had a high, negative correlation coefficient (R=-0.58) which indicates that these two categories assess similar data in an opposed way.
This approach of parallel scoring ultimately resulted in 7 selected candidate targets in month 18 and target-specific validation work has been carried out. Between month 18 and 36, continuous screening for novel candidate targets yielded 15 new targets that were reviewed between month 37 and 48 using a similar procedure, in an attempt to identify targets with even better properties. Between month 37 and 48, further validation activities took place, as well as acquisition of additional omics data that was used as supporting evidence.
• To further characterize available, and develop new in vitro and in vivo mechanistic and efficacy models that enable the monitoring of the biological function of CVD targets.
To maximally support mechanistic validation of the selected CVD drug targets, we defined the best suited models and strategies aimed at endothelial cell function, atherosclerotic plaque development and lipid metabolism. We developed both in vivo (rodent) models and in vitro (cellular) models that allow evaluation of pharmacological and mechanistic aspects of the target validation projects.
Novel in vivo models are being generated aiming for:
i. Plaque rupture rodent model.
Quorics previously identified several genes whose expression was correlated to endothelial cell dysfunction and leakage. These might provide an opportunity to generate a plaque rupture model, which would be the first such model in the CVD field and highly desired to support drug development projects aiming at treating atherosclerosis. Inducible endothelial specific expression models for ETS2 and PLVAP and the conditional THSD1 knockout have resulted in transgenic lines. To limit the generation of mice and to save costs and time involved in the in vivo testing of the new mouse models, we first decided to test the constructs on functionality in vitro using bone marrow (BM) differentiated endothelial cells. These tests allowed for a preselection of expressing constructs and selection of a small number of transgenic lines for further analysis. From these lines, by breeding with a Cre delete line, we generated ubiquitous deletions for THSD1, and ubiquitous activations via lox-stop-lox deletions, of PLVAP and ETS2. Currently, we are testing endothelial permeability on these animals. The actual usefulness of these models will have to be established by in depth characterization and functional analysis, and will determine whether another partner will continue experimenting on them, after the official end date of CarTarDis. To successfully characterize these models, partners in this WP have decided to extend their collaboration for the coming year.
ii. Aging rodent model.
We decided to start creating a controlled ageing model based on the expression of progerin as a translational tool for research of vascular ageing and other age-related diseases. In order to achieve a mouse model with regulated expression of progerin cDNA, we used the elements that were established in tissue culture for mifepristone-controlled gene expression. As this system is built on the transient or stable co-expression of a regulator and a responder plasmid, we had to combine and test a one-vector construct. Such a setup is more convenient for generating transgenic animals, but it poses a challenge for the correct dosage and expression strengths of the two components, and as such, had therefore to be tested in tissue culture. We ultimately decided to attempt generating transgenic animals with two different constructs that have an either low or strongly driven regulator element. Both constructs were microinjected and yielded transgenic founders. Currently, we are in the process of breeding offspring for testing uninduced and induced gene expression of progerin, and hope to obtain positive results right after the official end date of CarTarDis. Given the encouraging outcome of the LPL experiments, the prospects are high that this model is functional.
iii) Target-specific rodent models.
Of the 7 selected targets, Of the 6 selected targets, only two justified the generation of in vivo rodent models to support target validation: LPL and IL6R. Unfortunately, genomic evolutionary analysis showed that IL6R and its downstream pathway was insufficiently conserved from man to mouse. To generate a model with controllable downregulation of LPL, we targeted mouse ES cells with a construct that was validated in a similar way as described above, and that contained a duplicated operator site for the mphR(A) regulator about 250bp downstream of exon 1. These cells were used to generate chimeric animals via blastocyst injections, which led to germ-line stabilized heterozygotes. In order to exclude an influence of the operator insertion alone on LPL expression, the mice were bred to homozygosity and analyzed for phenotypic changes at PG, and further characterized by measuring LPL expression at Umea. We observed an odd and surprising tail phenotype, but otherwise the mice are healthy and LPL expression is unaffected. These mice were bred with a number of strains that were generated at PG, and express a KRAB-based suppressing regulator in different configurations. Currently, 4 co-transgenic models are being bred at PG. For three of them (17 animals in total), we have generated first pilot data by applying mifepristone via i.p. injections or in drinking water. These models utilize co-transgenics with a heterozygous LPL locus, and therefore, we expected a limited effect given the strong post-translational levelling of gene expression. Expectedly, LPL expression was uninformative in most tissues, while expression in kidney, where post-translational correction is reportedly absent, we observed a clear correspondence and full suppression (to 50%, as expected in the presence of a wild-type LPL copy). We can therefore anticipate that mice with a homozygous LPL-targeted copy can show a full response to regulation in all relevant tissues.
When applicable, we have to opportunity to combine the new models with the ApoE3Leiden mouse, as drug-responsive model for hyperlipidemia and atherosclerosis, and PCSK9 AAV transductions to rapidly increase plasma cholesterol levels and susceptibility for atherosclerosis development.
Opportunities for functional assessment of some selected targets using therapeutic agents such as small molecules, antibodies and recombinant proteins were carefully considered for each target. Our model of choice is the well validated E3Leiden.CETP model at TNO. To further validate LPL as a target for atherosclerosis we decided to perform in vivo studies using an LPL activating tool compound (LP071). LP071 has previously been shown to ameliorate the postprandial response after an olive oil gavage by lowering triacylglycerol (TG) in hypertriglyceridemic apolipoprotein A-V-deficient mice and in a short term study in the APOE*3Leiden.huCETP (E3L.CETP) mouse model. An atherosclerosis study was designed for E3L.CETP together with pilot experiments to investigate dosing and administration routes (IP and oral). However, the three pilot studies using different dosages and ways of administration did not show any effects of treatment compared to vehicle controls. Compound plasma exposure levels were investigated and found to be in the low µM range with clear distinction between oral and IP delivery and high vs low dosing. Since previous studies had been performed using a different vehicle this was also investigated in a separate pilot dosing study without significant changes in plasma lipids. The results were followed by repetition of previous studies on WT mice. Mice were treated with IP injections for 4 days and subsequently challenged by an oral fat tolerance test. However, after analysing the results it was clear that activation of brown adipose tissue erased any differences between the treatment groups. The study was subsequently repeated in thermo-neutral conditions, ablating any effect relating to thermogenesis, where a minor effect on postprandial lipid clearance and LPL tissue activity could be found in the LP071 treated group. The limited efficacy combined with the difficulties to perform an atherosclerosis study in thermo-neutral conditions prompted a stop with further experiments using this tool compound. Currently these results are investigated to determine the reason of this irreproducibility. The subsequent atherosclerosis study is put on hold until we have obtained more information. To further validate the role of OSM in endothelial activation, we proceeded with an in vivo study testing both purified OSM protein and inhibitory OSM antibodies as therapeutic agents. First, a pilot study was performed to determine optimal dosage and a potential in vivo effect using markers of endothelial activation (inflammation) and plasma lipids as read-outs. Subsequently, a full atherosclerosis study will follow evaluating the effect of OSM/OSM inhibition on atherosclerosis development. To support the functional target assessment on atherosclerosis we established several in vivo imaging methods including optimized protocols for echocardiography (3D and full cardiac scan, Doppler function of coronary and carotid arteries) and PET/CT (using 18-FDG, ongoing oral lipid tracer 18-FTHA and FA analogue 11-C acetate) as well as optical projection tomography protocols for detection of protein factors in aorta and heart.
In parallel, several in vitro cellular models to study endothelial cell and macrophage function have been developed, optimized and applied in hypothesis-driven target validation studies, guided by expertise in WP4. Initially, the developments of cellular models were focused on endothelial cells, since there was considerable interest and experiences within the consortium related to this specific cell type in CVD. Proper function of the endothelial cells is crucial for vascular maintenance and homeostasis. Dysfunction of the endothelial layer plays an integral part in both the onset as well as progression of atherosclerosis. Furthermore, during target selection, endothelial function was integrated in the capabilities and pathways criteria. Therefore, many of the selected targets are expressed in endothelial and/or affect their function. New models focussing on other CVD relevant cell types (VSMC, monocytes, neutrophils) have been developed to answer specific questions. Below is a summary of major findings related to the use of cellular models for validation of some of the selected candidate CVD targets.
i) Endothelial cell models.
We evaluated the role of two genes that are used to generate the plaque rupture model (PLVAP and TNFAip8L1) on several aspects of endothelial (dys)function. PLVAP was show to promote endothelial survival and migration by indirect stimulation of the production of the pro-angiogenic factor S1P via eEF1A1. We observed that TNFaip8L1 inhibits endothelial cell apoptosis through the inhibition of caspase 8 activity and subsequent cleavage of caspase 3. These findings were being used to prepare two manuscripts that describe the endothelial function of the targets. We also assessed the expression of the prioritized targets and relevant pathway markers in healthy and stimulated primary endothelial cells using RT-qPCR. So far, mRNA expression could be confirmed for the targets MCL1, PECAM-1, LRP, LIPA, PDGFD, ANGPLT1, ABCA8, PPAP2B, EDNRA and IL6R. CYP17A1 and CTF1 showed no mRNA expression in endothelial cells.
Previously we found that the target OSM induces endothelial activation in Human Umbilical Vein Endothelial cells (HUVECs) and murine endothelial cells. We now discovered that that OSM also induces endothelial activation in cells from different vascular beds, Human Aortic Endothelial Cells (HAECs) and Human Microvascular Endothelial Cell (HMEC-1) as characterized by higher MCP-1, IL-6 and ICAM-1 mRNA levels, increased ICAM-1 membrane expression and increased MCP-1, IL-6 and E-selectin protein release. In addition, we observed a marked STAT1 phosphorylation, indicating involvement of the JAK/STAT pathway in OSM signalling. siRNA knockdown of the LIFR and the OSMR revealed that simultaneous knockdown is necessary to significantly reduce MCP-1 and IL-6 secretion and STAT1 and STAT3 phosphorylation after OSM stimulation. These data are reported in the manuscript: Inflammatory cytokine oncostatin M induces endothelial activation in macro- and microvascular endothelial cells and in APOE*3Leiden.CETP mice (submitted). PPAP2B mRNA is expressed in human endothelial cells and silencing of the gene resulted in enhanced endothelial activation in previous experiments (Figure 10A). As PPAP2B was deemed not druggable we focused on its downstream LPA receptors that are functionally closely linked and are druggable. Interestingly, of the six known receptors for LPA, the target of PPAP2B, LPAR6, was observed to be the highest expressed on endothelial cells. This suggested that higher levels of LPA, due to the knockdown of PPAP2B may enhance endothelial activation primarily through LPAR6. To further explore this hypothesis, we next stimulated the HUVECs with LPA, what led to increased expression of cytokines IL6 and MCP-1 as well as ICAM-1 and VCAM-1. These effects were aggravated when PPAP2B was silenced, again suggesting that lower LPP3/PPAP2B levels may cause endothelial activation through enhanced LPA signalling. To test the hypothesis that the effect of LPA would be mediated via LPAR6, shown to be the most prominent LPAR expressed in ECs in human plaques, we continued with LPAR6 silencing. Interestingly, silencing of LPAR6 in HUVECs inhibited LPA-induced endothelial activation, observed by the reduced expression of ICAM-1, VCAM-1, IL6 and MCP-1. However, simultaneous silencing of both LPAR6 and PPAP2B did not result in any additional cumulative effects on endothelial activation compared to those already observed after silencing PPAP2B alone. Based on these findings, we speculate that, although silencing of PPAP2B and LPA stimulation induce similar phenotypes in ECs, they may not share the same LPA-LPAR6 signalling. Our results also indicate a complexity and flexibility in signalling patterns among LPARs, that can be dictated by repression of PPAP2B in combination with some yet unknown factors in the atherosclerotic tissue environment, considering that silencing of LPAR6 did not have any further effect on the endothelial activation already observed after silencing of PPAP2B alone. These data are described in the manuscript: Integrated human evaluation of the lysophosphatidic acid pathway as a novel therapeutic target in atherosclerosis (submitted).
ii) Immune cell, smooth muscle cells and fibroblast cell models.
Since some questions arising from within the consortium cannot be answered by the use of endothelial cells, we have developed functional human monocyte/macrophage, smooth muscle cell and fibroblast models in order to provide new tools. Based on data from the Karolinska BIKE biobank lower levels of PPAP2B are associated with the larger presence of immune cells and higher levels of several LPArs in human atherosclerotic lesions. Therefore, we evaluated if stimulation with LPA and/or knockdown of PPAP2B had an impact on the expression of the different LPArs on monocyte and macrophages. Both LPA and PPAP2B knockdown didn’t affect the expression of the LPARs on these cells in vitro. These data are incorporated in the manuscript on PPAP2B validation (noted above). To elucidate the trans-signalling effect of IL6r we stimulated monocytes, macrophages and fibroblasts with or without siRNA mediated knockdown of the IL6r in these cells. No clear effects were seen, which we believe are due to the high levels of IL6 that are produced by these cells. New experiments have been scheduled that take this into account.
iii) Foam cell models.
Biobank expression data from WP3 showed that many of our targets (OSM, PPAP2B, IL5 and CECR1 co-localize with foam cells. We ran experiments to turn monocytes/macrophages into foam cells, to study the role of these targets in these cells.
iv) Peripheral Blood Mononuclear Cells (PBMCs).
We also performed experiments using freshly isolated human PBMCs to elucidate the effect of CECR1 on immune cells. According to literature CECR1 should affect the immune status, by inducing the differentiation of immune cells towards an immune suppressive phenotype and subsequent induction of proliferation. Using flow cytometry, we could not identify an effect of CECR1 on B-cells and T-cells. Interestingly, in a pilot experiment we found an inverse relation between CECR1 levels and the number and activation of neutrophils, suggesting that neutrophils may be the real target of CECR1. Currently we are expanding the number of patients to substantiate this claim.
v) Neutrophils.
Recently the target CECR1 has been implicated to play a role in the activation of neutrophils. These changes in neutrophil biology are accompanied by compromised endothelial integrity, endothelial cellular activation, and enhanced inflammation. Together, this suggests that CECR1 may play role in atherosclerosis development by regulating neutrophil activation. To assess the effect of our target CECR1 on neutrophils in vitro, freshly harvested human neutrophils were stimulated with adenosine. Adenosine strongly induced neutrophil activation as seen by increased release of the atherosclerosis promoting factors MPO, MMP9, and neutrophil extracellular traps (NETs). Co-stimulation with CECR1 resulted in a reduction of MPO, MMP9, and NETs. These results suggests that CECR1 inhibits neutrophil activation by the conversion of adenosine to inosine. The subsequent effect of CECR1 inhibited activation on other cells and the process of atherosclerosis needs further investigation, which will continue post CarTarDis.
vi) Mural cell models (VSMCs).
To elucidate the seemingly complex role of IL6-IL6r signaling, we performed a series of in vitro experiments in which we silenced either IL6 and/or IL6r in various cell types. Thus far we observed no differences that suggest any differences in IL6 signaling. Since the production of IL6 in cultured cells is still relatively high even under knockdown condition, we decided to also include a neutralizing antibody, to really block the signaling cascade. These experiments are currently being performed and will continue post CarTarDis.
• To develop and optimize cutting edge molecular assays to accurately monitor proteins, RNA and metabolites in preclinical and clinical biosamples relevant to CVD.
- Biobanks:
To support the validation of candidate targets in human cardiovascular disease, we had access to two human biobanks containing well-annotated, appropriately stored, high quality tissue biosamples with high relevance to CVD. Socrates (LUMC) is based on material obtained during organ transplantation (coronary and aorta wall samples), and BiKE (KI) is focussed on surgical specimens, in particular carotid endarterectomies. Socrates (LUMC) is largely based on perirenal aortic patches (from over 500 donors) that are collected during kidney transplant. Unlike other atherosclerotic biobanks, which are all based on end-stage surgical specimens or post mortem samples, material in Socrates is from relatively healthy individuals who fulfilled the Eurotranplant criteria for kidney donation. As such material in Socrates covers the full life-span (5-80 years of age), as well as the full spectrum of atherosclerotic disease. Removal of the aortic patch along with the renal artery is a standard procedure in order to preserve the integrity of the renal artery of the donor graft. At the time of transplantation the aortic patch is removed, split in half with one half frozen and the other half formalin fixed. A Movat staining is performed for each section and the stage of atherosclerosis graded (adapted AHA classification as proposed by Virmani). Material from the Socrates biobank was exchanged between partners of WP3 and used extensively to characterise expression of the selected targets and associated pathway biomarkers in the tissue samples representing various phases of CVD. The BiKE resource (KI) contains biosamples from patients who are scheduled for surgery for symptomatic or asymptomatic carotid stenosis and are included prospectively after written informed consent. After surgical retrieval, the endarterectomy/plaque is rinsed in physiological buffer and divided transversely at the culprit of the lesion. The proximal part is snap frozen in liquid nitrogen and the distal part fixed in formaldehyde (or embedded in O.C.T. compound and frozen). Blood samples are obtained simultaneously for plasma preparation and isolation of PBMCs for genotyping and gene expression analysis. Plaque tissue is utilized for isolation and preparation of mRNA, miRNA and cDNA. Furthermore through the highly optimized workflow, BiKE allows a rapid and efficient identification and validation of potential target candidates, incorporating transcript profiles, in situ protein detection and correlations with clinical data, genotypes, plasma levels and morphology. Both LUMC and Karolinska continued to update their human biobanks. Sampling procedures were adapted in order to collect samples suitable for further downstream analyses, which include immunohistochemistry, mRNA and microRNA detection, and lipid mass spectrometry. Finally, a preclinical biobank was set up at TNO that mirrors the human biobanks and enabled cross-species molecular analysis of the selected CVD targets and their pathways.
- Sample and data exchange:
The consortium partners developed a multi-partner sample exchange program whereby the same CVD tissue sample was analysed by multiple partners with their specific molecular staining methodologies. Protocols for analysis of plaque samples were optimized by the different partners and in parallel, best practice knowledge on staining protocols was exchanged. This workflow was used for the prioritized CVD targets’ studies. Snap frozen human carotid plaques (from KI) were sent to Imabiotech for lipid and metabolite imaging by Mass Spectrometry Imaging (MSI). Sections from the same plaques were sent back to KI for Immunohistochemistry (IHC) imaging and to Bioneer for mRNA In situ hybridisation (ISH) imaging. MSI, IHC and ISH pictures were shared between partners. Since routine tissue processing techniques interfere with MS-imaging application Imabiotech relies on dedicated native tissue samples for its mass spectrometry imaging. As this limits the application of MSI, Imabiotech has investigated methods to use formalin-fixed tissue for MS-imaging. To enable data sharing among CarTarDis consortium partners, Morphisto has established a virtual microscopy environment that allows on line data sharing. Imabiotech developed a server-based data storage and exchange facility that allows easy access with partners. LUMC developed software to share IHC images obtained using their Philips digital microscopy system operative.
- Innovation in Molecular staining methodologies:
The partners in the CarTarDis consortium WP3 further expanded their expertise in molecular staining methodologies. For staining of mRNA and miRNA, an automated in situ hybridization (ISH) platform was established by Bioneer using branched DNA probe technology (RNAscope) halfway the project. Also, Bioneer developed a double staining method based on fluorescence to visualize mRNAs and cellular markers in the same atherosclerotic tissue, e.g. to document the expression of PPAP2B and LPAR6 in luminal endothelial CD34+ cells in the plaque. Using these methods, Bioneer completed the optimized screening and evaluation of the expression of mRNA from the first selection of candidate target genes: OSM, OSMR, LIFR, CECR1, IL5, IL6R, PPAP2B/LPP3, LPAR1, LPAR2, LPAR5 and LPAR6 in eight FFPE-blocks from the Socrates biobank representing the following stages: AIT, IX, PIT, EFA, and LFA of atherosclerosis development. The complete screening was performed using the brown-staining protocol using DAB, and a subset of the FFPE-blocks representing stages IX, EFA and LFA were subsequently stained using the red dye FastRed. The use of the red dye, enables combining the mRNA-ISH signal with immunofluorescent staining of cells to provide evidence of the origin of the cell expressing the genes.
For staining of proteins, Morphisto optimized immunohistochemistry (IHC) methods for the selected targets and relevant pathway protein biomarkers and receptors, in good alignment with the IHC development at Karolinska and LUMC. In addition, Morphisto optimised classical histochemical staining methods and improved throughput of their staining workflow. Staining tasks for the respective targets was divided among the partners, whereas in addition, Morphisto executed expression profiling of the top 20 targets in human lesions from the Socrates biobank (LUMC) that spanned the full cardiovascular disease spectrum. Double and multiple IHC protocols were developed by KI and LUMC to investigate and validate the cell type specific expression and localization of the different targets. LUMC developed and validated a versatile protocol that allows for sequential staining cycles on a single sample, thereby allowing simultaneous appreciation of multiple markers (IHC analysis of 7 protein biomarkers was demonstrated).
For staining of small molecules including lipids, Imabiotech applied their mass spectrometry imaging (MSI) platform to semi-quantitate specific molecules such as cholesterol esters (CE), Lysophosphatidic acid (LPA), phosphatidylcholine (PC), phosphatidic acid (PA) and others. These were particularly used to establish co-localization of selected lipid substrates and aspects of the PPAP-2B pathway. In addition, Imabiotech developed specific assays to measure and localise tissue levels of adenosin/inosin in atherosclerotic samples from the Socrates biobank that were linked to expression levels of CECR1 mRNA and protein. To further innovate their MSI capabilities, Imabiotech developed methods to use formalin fixed tissue for MS imaging, thus expanding the application of their MSI platform to evaluation of non-surgical specimens.
For analysis of gene expression in specific tissue regions, KI has optimized laser capture microdissection (LCM) microscopy for isolation of specific human carotid plaque regions (fibrous cap, shoulder region and necrotic core). LCM is a technology that allows isolation of single cell populations under direct microscopic visualization, followed by various downstream applications as RNA transcript profiling, cDNA library generation, proteomic studies and discovery of novel signalling pathways. By using LCM, we have successfully isolated α-actin positive cells (αSMA+) in the fibrous cap of human endarterectomies. RNA quality and concentration were evaluated after every LCM by bioanalyzer and RNA integrity number (RIN) was estimated to be between 3 and 6. RNA concentration was approximately 0.3-3 ng/μl. We demonstrated that analysis of expression of housekeeping genes (GAPDH, RPLP0) by quantitative RT-PCR (Taqman) showed a RNA concentration dependent curve, indicating the LCM-gene expression assay is ready for analysis of specific target genes.
- Application to validation of selected targets:
We successfully applied complementary techniques to study the prioritized CVD targets identified in WP1 and WP4. A combination of metabolic data from WP1 (HMGU) and genetic analysis (Icelandic Heart Association, IHA) suggested an association between lysophosphatidic acid species and incident of plaque development. Previously, using samples from the BiKE biobank, ImaBiotech showed an accumulation of lysophospholipids in the shoulder of the plaque. A combination of lipid imaging together with specific IHC (from KI) identified the overlay regions in the plaques between lysophospholipids and lipid phosphate phosphohydrolase 3 (LPP3/PPAP2B). Along similar lines Karolinska, LUMC, Bioneer and Morphisto exchanged samples in order to validate the prioritised targets by ISH and IHC. The successful integration of complementary techniques to study the prioritized CVD targets is illustrated by the work on PPAP2B, the lysophosphatidic acid (LPA) pathway and atherosclerosis susceptibility. Based on the genetic data from IHA, the PPAP2B gene encoding for lipid phosphate phosphohydrolase 3 (LPP3) was prioritized as one of the primary targets in the project. Using samples from the BiKE biobank, ImaBiotech showed an accumulation of lysophospholipids in the shoulder of the plaque. A combination of lipid imaging together with specific IHC (from KI) identified the overlay regions in the plaques between lysophospholipids and lipid phosphate phosphohydrolase 3 (LPP3/PPAP2B). A similar combination of techniques was applied to detect LPAR6 protein by immunohistochemistry staining (KI) and LPAR6 mRNA by in situ hybridization (Bioneer) in human atherosclerotic tissue. Data generated are intrinsic part of all of the target validation projects and of the corresponding manuscripts that have been prepared.
During the last months of the CarTarDis project, Bioneer and Morphisto undertook the task of screening the entire set of top 20 target genes defined in WP1 and WP4 on mRNA and protein level respectively. The samples were provided by the Socrates biobank. After initial screening of the extra blocks a set of nine blocks were selected for the mRNA-ISH screening of the 20 targets. At Bioneer, 20 genes were screened using the automated staining platform and the RNAscope branched-DNA probe technology with red staining. All slides were scanned and evaluated for mRNA expression in different compartments (adventitia, media or intima) of the different stages of atherosclerotic plaques. Additionally, ISH-screening of OSM, OSMR and LIFR on the same nine samples mentioned above was performed to support the studies and publications on OSM. Together with KI, Quorics, LUMC and ImaBiotech a multimodality imaging and staining study was also established using a cryo-preserved carotid plaque for BiKe. CECR1-ISH, parallel CECR-IHC and MSI to image the presence of inosine and adenosine was performed on serial sections prepared and distributed by ImaBiotech for multimodal imaging. Similarly, a multimodality imaging and staining study on another set of serial sections, involving CD31/PECAM-ISH, PLA2-IHC and MSI to image LPA and LPC phospholipid species was also done to support the technology focused paper and poster to be presented at the AHA-2017 meeting.
• To apply molecular assays to further characterize biosamples of two comprehensive well-annotated human CVD biobanks to investigate and validate the contribution of specific pathways to CVD.
Analysis of the two unique human biobanks in the CarTarDis consortium (Socrates, Bike) that contain precious biobanked human material from CVD patients in all stages of disease has provided insightful data to support target validation including co-expression of the selected targets and associated pathways and evidence for their clinical relevance. The section above outlines how the novel molecular staining protocols were applied to human CVD biosamples from the Socrates and Bike biobanks in the development and validation of the methods and of the evaluation of the preselected candidate drug targets. In the final year of the CarTarDis project, the implemented and developed methods were used to visualise the top 20 targets selected in our workflow using samples from the Bike and Socrates biobanks.
• To develop and exploit this newly formed pan-European modular pharmaceutical target discovery network in cardiovascular disease.
The intense collaboration between the CarTarDis consortium partners during the development of novel tools and models and during the validation studies of the selected targets provided a network among the multidisciplinary experts as originally planned. This has yielded several spin-off collaborations between some partners, whereby the unique expertise of a partner contributed to contract research projects of other partners (details not to be disclosed due to confidentiality) or whereby partners involve each other in related activities (eg close interaction between biomarker networks of Netherlands (prof A van Gool) and Denmark (dr K. Holmstrøm). In addition, the joint consortium meeting with the three FP7 CVD consortia widened the window of opportunity for collaboration with other peer scientists. This yielded several bilateral collaborations in contract research, joint publications of opinion papers and submission of novel joint research proposals. The topics of these novel interactions include cardiovascular research, translational biomarker development and application of molecular technologies in translational research.
Potential Impact:
Guided by the nature of the EU FP7 call (SME targeted ambitions) we defined the overall objectives of the CarTarDis project as: 1. Identify and validate novel CVD drug targets, 2. Develop in vitro and in vivo tools to support CVD drug development, 3. Increase the knowledge and competitive position of the participating partners (notably the SME’s) in drug discovery. The potential impact is best to be described for these three categories.
i) Novel CVD drug targets.
Directed by the experienced pharmaceutical scientists in our team and advised by experienced drug developers and entrepreneurs in our External Advisory Board, we designed a rational approach to accurately reflect the pharmaceutical process, adopting best practices from pharmaceutical industry from its start. The selection filters were inspired by the 5R strategy of AstraZeneca and included Genetic/omic evidence, Druggability, Novelty, Feasibility and Pathways. A database and target selection workflow was designed that allowed us to select and prioritise candidate CVD drug targets. For each preselected target, a target champion was appointed who led a multidisciplinary team of consortium scientists to define the key experiments to generate data to support the validation of that particular target. All teams were guided and supported by the experienced pharmaceutical drug hunter scientists and clinical cardiovascular experts in our consortium, together forming strong teams to reach the objective of identifying novel validated cardiovascular drug targets. First, available and emerging molecular data from three large independent cardiovascular cohorts was mined for statistical association with cardiovascular disease phenotypes, with a preference for genetic associations. Then, validation of the candidate targets was done by two parallel approaches, in which mechanistic validation was performed using novel in vitro (cellular) and in vivo (rodent) preclinical models in which the candidate targets were studied regarding mechanism of action and role in particular aspects of the cardiovascular processes and clinical validation was based on two unique human cardiovascular tissue biobanks and several high resolution molecular analysis methods using which the spatial expression of the target and target-related pathways in cardiovascular tissue samples was studied on RNA, protein and small molecule level. Of the 381 preselected candidate targets, we prioritised 7 targets, i.e. PPAP2B, OSM, CECR1, LPL, IL6R, IL5, ApoBD. Seven multidisciplinary target validation teams were formed involving all partners of the CarTarDis consortium. In line with current practice in pharmaceutical drug discovery, we experienced the complexity of human biology and the new data supported further validation for only 3 of the selected targets (PPAP2B, OSM, CECR1). For some, we identified downstream proteins as potentially better drug targets than the originally preselected target and focussed our validation experiments on those (for instance LPAR6 was a more druggable target whereas the initial hit was PPAP2B. We sought interaction with parties for identification of potential adopters of the validated drug targets from the CarTarDis project for subsequent drug development. These included (semi-)industrial partners in our network, notably AZ and the European Lead Factory, and our External Advisory Board, notably Menzo Havinga and Erik Lund. This did not lead to start of an adoption process before the close of the CarTarDis project. Although we progressed significantly, the validation status of the selected targets was not at a level yet to merit subsequent investments by private partners. We finished the validation projects with clear view on additional experiments that are needed beyond the project to provide ultimate support for the chosen targets. All studies will be published to impact the scientific cardiovascular field, using peer-reviewed publications and posters at CVD conferences (such as the recent American Heart Association 2017).
ii) in vitro and in vivo tools to support CVD drug development.
The partners in the CarTarDis consortium developed novel innovative tools that support the validation of candidate drug targets in cardiovascular disease and other therapeutic areas, and support the development of drugs targeting those novel targets. Among the ICT tools that were developed, the interactive database that we used to prioritise the preselected CVD drug targets (developed by HMGU) received much appreciation from our External Advisory Board, outlining that such approach was rarely been followed in the academic communities and would be worthwhile disseminating (manuscript is in preparation). The data exchange solutions developed by consortium partners in (notably LUMC, Morphisto, Bioneer, Imabiotech) will facilitate future interactions of these partners in bilateral and multi-partner collaborations. The in vitro (cellular) models and in vivo (rodent) models that were developed (detailed above) will impact the cardiovascular research field as they provide new capabilities to translational research, either using the developed assays or methodologies. To illustrate, the novel targeting strategy using to generate a tissue-specific expression modulation of LPL (developed by Polygene) will be applicable to many more rodent models in the future (patent filed). The functional cardiovascular cellular models in combination with siRNA mediated knock-down of gene expression used to study the mode of action of specific genes, will not only impact the cardiovascular researchers but also provide a template workflow that can be used by other researchers (multiple publications published or in preparation). The molecular staining methodologies are described in explicit detail above. Notably, the mass spectrometry imaging technology to highly specific visualize particular small molecules in tissue slides (Imabiotech) was rather unknown in the cardiovascular field, whereas we could clearly demonstrate the added value of this approach, particularly in combination with other imaging modalities (multiple posters and papers have been prepared). The branched DNA approach to visualize specific mRNA transcripts in tissues in an amplified manner (developed by Bioneer) has the potential to significantly impact molecular analysis of tissues by increasing the sensitivity of spatial gene expression analysis.
iii) Competitive position of the participating partners (notably the SME’s) in drug discovery.
The SME’s in the CarTarDis consortium participated strongly in these two validation activities by contributing their unique technological expertise as part of the validation teams. Throughout the project, we paid particularly attention to the educational aspect of the pharmaceutical process by outlining the rationale behind particular decisions in the pharmaceutical pipeline and involving all partners in this process. Judged by the feedback of the partners (particular the SME’s) during the project and upon finishing the final period, this was much appreciated. Particularly the knowledge of participating partners with still limited experience in the pharmaceutical process was boosted to an experienced level that impact their ability to interact with pharmaceutical researchers in the future. The joint consortium meeting with the three FP7 CVD consortia (Sept 2016) increased the exposure of the partners and widened the window of opportunity for collaboration with other peer scientists. This yielded several bilateral collaborations in contract research, joint publications of opinion papers and submission of novel joint research proposals. The topics of these novel interactions include cardiovascular research, translational biomarker development and application of molecular technologies in translational research. As such, the CarTarDis consortium members have become more attractive collaboration partners in bilateral and multi-partner collaborations, one of the key objectives of the CarTarDis project.
List of Websites:
http://cartardis.eu
The EC FP7 funded project CarTarDis (Cardiovascular Target Discovery) aimed to apply the high quality expertise and resources within its multi-partner consortium to discover novel therapeutic drug targets with a high level of clinical correlation to cardiovascular disease (CVD). The project adopted a strong pharmaceutical work flow in which the 13 consortium partners formed a virtual pharma company in which human cardiovascular cohorts, preclinical models, clinical biobanks and innovative molecular staining methods played a key role. The overall objectives of the project were three-fold: 1. Identify and validate novel CVD drug targets, 2. Develop in vitro and in vivo tools to support CVD drug development, 3. Increase the knowledge and competitive position of the participating partners (notably the SME’s) in drug discovery.
Directed by the experienced pharmaceutical scientists in our team and advised by experienced drug developers and entrepreneurs in our External Advisory Board, we designed a rational approach to accurately reflect the pharmaceutical process, adopting best practices from pharmaceutical industry from its start. First, available and emerging molecular data from three large independent cardiovascular cohorts was mined for statistical association with cardiovascular disease phenotypes, with a preference for genetic associations. A database and target selection workflow was designed that allowed us to select and prioritise candidate CVD drug targets. The selection filters were inspired by the 5R strategy of AstraZeneca and included Genetic/omic evidence, Druggability, Novelty, Feasibility and Pathways. For each preselected target, a target champion was appointed who led a multidisciplinary team of consortium scientists to define the key experiments to generate data to support the validation of that particular target. All teams were guided and supported by the experienced pharmaceutical drug hunter scientists and clinical cardiovascular experts in our consortium, together forming strong teams to reach the objective of identifying novel validated cardiovascular drug targets. Validation of the candidate targets was done by two parallel approaches. Mechanistic validation was performed using novel in vitro (cellular) and in vivo (rodent) preclinical models in which the candidate targets were studied regarding mechanism of action and role in particular aspects of the cardiovascular processes. Clinical validation was based on two unique human cardiovascular tissue biobanks and several high resolution molecular analysis methods using which the spatial expression of the target and target-related pathways in cardiovascular tissue samples was studied on RNA, protein and small molecule level. Particularly the combination and high resolution of the applied technologies was unique. The SME’s in the CarTarDis consortium participated strongly in these two validation activities by contributing their unique technological expertise as part of the validation teams.
The CarTarDis project was successful in several aspects but not in all, in line with the state of the art in pharmaceutical industry. More specifically, we designed a unique interactive database that allowed us not only to summarize key features of preselected candidate drug targets but also to easily adapt the selection criteria to yield a mixed portfolio of drug targets varying in novelty, druggability, omics-based evidence and scientific value. Of the 381 preselected candidate targets, we prioritised 7 targets and formed 7 multidisciplinary target validation teams involving all partners of the CarTarDis consortium. We designed novel in vitro (cellular) and in vivo (rodent) models and multiple novel molecular staining methods that supported the validation studies but also provided added value to the expertise of the consortium partners and to the biomedical CVD field at large. In line with current practice in pharmaceutical drug discovery, we experienced the complexity of human biology and the new data supported further validation for only 3 of the selected targets. For some, we identified downstream proteins as potentially better drug targets than the originally preselected target and focussed our validation experiments on those. We finished the validation projects with clear view on additional experiments that are needed beyond the project to provide ultimate support for the chose targets. Overall, the educational aspect of this process was highly valuable for the participating partners with still limited experience in the pharmaceutical process, one of the key objectives of the CarTarDis project.
Project Context and Objectives:
The development of a novel drug is an extremely challenging task. On average and across various therapeutic areas, the attrition during clinical development is 90%, meaning that only one in ten projects being pursued through clinical testing in patients is successful. The major cause for this high attrition is the proof-of-concept phase, during which the attrition rate is as high as 80% because of a lack of efficacy and/or unacceptable safety liabilities. This leads one to surmise that preclinical studies in pharmaceutical research are insufficient to predict drug action in patients and that high clinical validation of the drug target is key to drug development success. As for other complex chronic human diseases, cardiovascular disease (CVD) is a difficult area to develop pharmaceutical drugs. CVD affect the heart and surrounding blood vessels and can take many forms, such as high blood pressure, coronary artery disease, heart disease and stroke. CVD is a growing medical problem globally with a predicted 23.6 million deaths from CVD in 2030, increased from a 17.3 million death from CVD in 2008 which represented 30% of all global deaths at that time. In the EU, CVD are the largest cause of death in the EU and account for approximately 40% of deaths or 2 million deaths per year. The slow development of the disease (which becomes overt in general above 50 years of age in man and above 60 years of ages in women but may start already at age of 20 ) and the limited availability of clinical samples and translational models has hampered the identification of drug targets and development of efficient therapies. In many cases, cardiovascular association studies have focussed on identifying underlying mechanism of disease and/or clinically relevant biomarkers but not on identification of novel drug targets, and consequently a large part of the obtained data is unsuitable for that purpose. Moreover, clinical material is lacking either in quality or in number of patients, and preclinical models to mechanistically investigate candidate targets are insufficiently translational to the clinic. To increase the chance for successful clinical development of novel therapies in CVD, further innovation is required by combining: 1. Independent and large-scale population studies from which novel targets with strong correlation to clinical phenotypes are deduced, 2. High quality human biobanks to confirm molecular relevance of targets in diseased cardiovascular tissues, 3. Translational disease-mimicking models (cellular and animal) to validate novel drug targets, 4. A stringent pharmaceutical drug discovery process for prioritization of druggable targets. Particularly the druggability assessment of a candidate target, e.g. whether a target protein contains a binding cleft for a drug (be it a small molecule, peptide or protein), is an imperative step in the pharmaceutical process which is rarely being done in clinical association studies.
The CarTarDis consortium brought together the required components mentioned above by careful selection of consortium partners that have a distinct expertise in cardiovascular pharmaceutical drug development and in preclinical and clinical cardiovascular research, in combination with unique resources such as cardiovascular cohort study data, clinical biobanks and molecular analysis methodologies. The required pharmaceutical focus at each step of the project was implemented using the strong pharmaceutical drug discovery experience in CVD among scientists working at the consortium partners. An important concept, in line with the changed pharmaceutical business model, is that these steps can best be executed by a functional network of relevant industry, clinical academic centres and specialized SMEs that brings together the right components and expertise to execute such modular target discovery process. The CarTarDis consortium operationalized a consortium of high quality academic institutions and leading SMEs with key expertise in cardiovascular disease and relevant technologies. It was composed of these parties that each have unique assets and expertise that only in combination can complete the objectives as defined in the project, and pursue the further development and exploitation of its output.
The CarTarDis project objectives were:
• To identify and select candidate CVD drug targets with a strong relevance to clinical phenotypes based on molecular and omics data from three unique large-scale population cohorts.
• To further characterize available, and develop new in vitro and in vivo mechanistic and efficacy models that enable the monitoring of the biological function of CVD targets.
• To develop and optimize cutting edge molecular assays to accurately monitor proteins, RNA and metabolites in preclinical and clinical biosamples relevant to CVD.
• To apply molecular assays to further characterize biosamples of two comprehensive well-annotated human CVD biobanks to investigate and validate the contribution of specific pathways to CVD.
• To apply the in vitro and in vivo models and human biobanks to prioritise and select CVD targets.
• To develop and exploit this newly formed pan-European modular pharmaceutical target discovery network in cardiovascular disease.
Project Results:
The CarTarDis project started 1 October 2013 and ended 30 September 2017. The consortium has disseminated the results and outcomes through presentations, scientific and popular articles and posters. CarTarDis built the foundations for successful development and positioning of emerging drug development tools and molecular methods into cardiovascular research. It also has strongly increased the knowledge of the participating partners in the process of pharmaceutical research and development, providing them a better competitive position in the translational biomedical science field.
• To design a robust target identification and prioritization workflow and criteria based on best practice in pharmaceutical drug development.
The consortium generated a desired workflow with application of filters and criteria based on the most relevant information available of the candidate gene targets identified and linked them to atherosclerosis and cardiovascular events. The workflow included the use of filters that should allow selection of targets for validation in humans to increase the confidence and reduce the risk associated with target selection. Because the biochemical characteristics of individual targets and their relation to different stages of disease development may differ, each candidate target will need to be individually evaluated according to described criteria and specific target-nature characteristics. This assessment includes evaluation of the target causal relation to atherosclerotic cardiovascular disease (ACVD) and if its modulation could be feasible by either small molecular weight compounds, peptides, antisense oligonucleotide or biologicals, such as a proteins or antibodies in a clinical setting (i.e.druggability). We adopted the 5R strategy designed and used by AstraZeneca, as this strategy has shown to be valuable in improving the success of a drug development pipeline. The 5R strategy outlines the assessment of drugs at various levels: Right Target, Right Tissue, Right Safety, Right Patients, Right Commercial Potential. Subsequently, we defined various levels of assessment of our portfolio of candidate CVD drug targets: Druggability, Novelty, Genetic/omic evidence, Feasibility, Pathways. The AZ Pharmconnect database was used to more accurately assess the “druggability” potential of each target. Druggability is a term used in drug discovery to describe a biological target (such as a protein) that is known to or is predicted to bind with high affinity to a drug. Furthermore by definition, the binding of the drug to a druggable target must alter the function of the target with therapeutic benefit to the patient. The concept of druggability is often restricted to low molecular weight organic substances but it has been extended to include large chemical molecules (anti-sense oligonucleotides) and biological molecules such as monoclonal antibodies. Sources of information for this approach included databases of curated drug targets, target screening databases (ChEMBL, BindingDB), literature (PubChem) and manually compiled sets of 3D structure known by the developers to be druggable.
Crucial to the selection of candidate drug targets from the wealth of information collected in the interactive CarTarDis database is the weighing of the selection criteria. Initially we regarded assayability and novelty as the major criteria for target selection, based on the knowledge at that time. However, recent years have seen a dramatic increase in therapeutic opportunities using novel therapeutic agents as RNA interference, small high affinity biological agents and so forth, making the assayability criterium less important. Hence, we adapted the weighing of the selection criteria such to yield a variety of targets with different characteristics and potential.
• To setup a database for all targets and make it available to all the partners.
We have set up a database server for storing of potential CVD drug targets and their annotation data. The system consists of a Unix based server, an SQLite database, and a web frontend based on Python (Django) and Perl. The SQLite database consists of four logical units: first, the target table, which represents the centre of the database; second, several tables to store relevant information on the targets, subdivided into categories, such as genetic information and animal model based information, amongst others; third, the user/institute based part, which allows tracking of the entries in the database; and fourth, a commenting and scoring system to enable a proper prioritization and evaluation of each target, between different partners within the consortium. The accessible web interface for storing and ranking of the potential drug targets was upgraded to adapt the criterion of data protection. Confidential information and the ranking of the targets was access restricted with a password (using the Apache-built-in ‘basic authentification’) and by a twofold IP check (once Apache-based, once interface-based). Additionally, an access-restricted website (https://sites.google.com/site/cartardistargetvalidation/) was set up as knowledge management platform to capture detailed insight into the pre-selected and prioritized targets. Each of the highly emerging targets has been intensively discussed, with ‘pros’ and ‘cons’, underlying mechanisms, druggability potential and other features. Target validation plans, activities and data summaries are shared through this website. This website is frequently updated with the latest available information, with input from multiple consortium members.
We initially collected 381 potential drug targets that were identified either in the public domain (CVD and lipid related GWAS data), or in studies (tissue-based work, animal models, genetic studies) internal to member sites within this consortium. Most of the targets were selected based on genetic evidence (188 associated with CVD and 157 with lipid levels). In general, the targets can be assigned to three groups. (1) high confidence targets at genomic sites linked to CVD in many cohort studies, (2) medium confidence targets showing suggestive evidence for association to CVD, most often in a single study, and finally (3) targets still to be evaluated for genetic links to CVD, but nominated for other reasons including prior art work on human tissues and animal models. The database integrated data from various databases and other sources to provide a comprehensive insight into the potential usability of the collected genes as drug targets. Considering the suggestions of the partners and following the decisions made at the consortium meeting in Gothenburg, we adapted the database to the specific needs of the prioritization for CVD drug targets. This mainly consisted of the commonly agreed relevant data that should be integrated and how it should be represented in the database to provide an informed overview of the targets. Examples of this included the integration of the expert scoring and weighing of the various criteria for drug target assessment. We provided systems level information for each target in order to enhance confidence in the target, provided insights into direction of the effect, uncover disease mechanism(s) and highlight pathways of interest. For this we have carried out an integrative target-specific approach based on various external data sources (e.g. online databases such as KEGG, the Gene Ontology, Ingenuity, Mouse Genome Informatics, the HMGU GWAS database, USCS, Orpha.Net OMIM, patent databases, DrugBank, ChEMBL, Proteinatlas amongst others) as well as internal data sources of human tissue work (molecular networks, regulatory variants like eSNPs/eQTLs, gene expression clinical trait correlations) and genetic data (exome- and genome-wide SNV/SNP data) analysed in a deeply phenotyped cohort (e.g. 13,000 with extensive plaque data). These types of analyses were carried out in both an automated fashion and manual curation for proper data mining. To enable the exploration of the targets with respect to only one of the predefined categories or a customized collection of them, we additionally implemented a filter to only display the categories of interest and fade out the others. This feature is especially helpful in terms of identifying knockout criteria and targets exclusion. Subsequent screenings of emerging literature and updated databases yielded more candidate drug targets that were included in the databases and compared with earlier selections to assess whether prioritisation should be changed. For example, we added 181 genes, which are located in the flanking area 200 kb up- and downstream of 57 SNPs with genome wide significant association to CAD, 511 genes from the CADgene database, 197 genes with blood eQTL values. Moreover, further 756 genes were subjected to our analysis pipeline of which 637 genes were previously suggested to be potentially involved in the development of CVD and 119 genes associated with Alzheimer’s disease that acted as a validation set. The results of our analysis showed that particularly genes from the CAD gene database (average score = 4.55) and those with an eQTL association (average score = 5.00) scored clearly better than those associated with Alzheimer’s disease (average score = 3.89) which is a clear indication that our approach is specific for CVD.
• To identify and select candidate CVD drug targets with a strong relevance to clinical phenotypes based on molecular and omics data from three unique large-scale population cohorts.
We initially pre-selected 381 highly informed potential drug targets from public and proprietary data sources and linked various levels of information to each, including those related to biological function and relation to CVD. We grouped the gathered information into the aforementioned five different categories of relevance. Then, we subjected the information of the candidate drug targets to two approaches of evaluation: an expert-based scoring and an automatic/systematic scoring. Expert scoring was done by experts in the CarTarDis consortium whose appreciation was quantitated and included in the CarTarDis database. However, given the continuous increase in published candidate CVD drug targets and the highly time consuming nature of expert scoring, we aimed to develop a systematic scoring system that would provide an automatic scoring, taking all aspects into account with predefined criteria, weighing of parameters and easy visualisation, that would allow us to test the contribution of criteria filters in an easy yet comprehensive manner. Hence, in parallel with the expert scoring of the first 381 targets, we developed, validated and improved a systematic scoring system. The latter was applied to newly entered candidate targets in the CarTarDis database to assess whether we had been selecting the targets with highest potential.
Expert teams were formed to define evaluation criteria for each of the six defined categories, namely OMICs evidence, novelty, druggability, involvement in lipid metabolism, endothelial dysfunction and inflammation. The expert teams were composed of experts in each domain within our consortium, and asked to apply these criteria to each of the categories and to manually assign scores to every target. The scaling ranged hereby from 0 (no confidence target) to 6 (high confidence) for OMICs evidence and from 0 (no confidence target) to 3 (high confidence) for all other categories. The double weighting of the OMICs evidence was chosen to emphasize the aspired association with the endpoint. In parallel, we attempted to define simple and significant rules that can be applied to easily minable and publically accessible data in order to systematically reproduce the values of such an expert scoring.
With respect to novelty, the experts investigated for each gene if there are patents, in particular with a CVD background, they took a look at the history of investigation a gene underwent, and if it is subject to current research. Our approach to implement this into a systematic score was to mine the PubMed database and extract the number of publications for each gene and specific key-terms. These five key terms were “atherosclerosis”, “coronary”, “cardiovascular”, “hypertension”, and “infarction”. Each of the terms was combined with the gene name and the term ‘gene’ and five scores (one for each term) were calculated. Genes with less than two publications available were scored with 3, genes with less than ten publications were scored with 2 and genes with less than 100 publications were scored with 1. In addition to these five scores, the publication count without the key-term was calculated. Hereby genes with less than 50 publications were scored 3, those with less than 150 were scored with 2 and those with less than 1000 publications were scored with 1. The final automatic score was finally calculated as average of the six previously derived scores.
Concerning the three relevant pathways (endothelial dysfunction, lipid metabolism, and inflammation) the expert scores were mainly based on online literature and databases. In order to automatize this process, we defined specific key terms for each of the three categories and checked their presence in the mined information from PantherDB and Kegg. Detailed information about the used terms can be seen in Table 1.
The experts for the genetic scoring considered several aspects: The presence of a genome-wide significant SNP for CAD, if any genetic variant of this gene is linked to a CVD-related phenotype, a connection to CAC or plaque, and transcriptomic observations. Our systematic approach to implement such a scoring was mainly based on the presence of a known CAD, CAC, MI or CHD associated SNP 200kB up- or downstream. If the significance of the SNP was genome wide (P<5E-08), the assigned score was 3, if the significance was P<1E-05 the given score was 2 and associations with P<1E-04 were given a score of 1. Additionally, an assignment in literature to a SNP with a genome wide significant association to blood lipid levels increased the score by 2 and a genome wide significant blood eQTL value with any of the CAD associated SNPs increased the score by 1 (Table 1).
Finally, the expert score for the druggability integrated complex considerations that were already subject to the scores for the other five categories, as well as tissue distribution of a specific gene, involvement of coagulation, lipolysis, amongst others. Consequently, the implementation of the systematic approach to adequately reflect these aspects was the most challenging, compared to the other systematic scores. To address the link to the endpoint, we used the information on known diseases associated with each of the 321 genes and increased the scores by 1 if any of the diseases was CVD. Furthermore, if a disease with an involvement of blood lipid levels was found, the score was also increased by 1. With respect to the classical definition of druggability- focussing on the question if the gene can be addressed with a small molecule drug – we evaluated the information on the protein type and protein location and increased the scores by 2 if the transcribed protein was located in the extracellular space or by 1 if it was located in the plasma membrane. Additionally, if the genes encoded G-protein coupled receptors, transmembrane receptors, growth factors, transporters, ligand-dependent nuclear receptors, kinases, cytokines, enzymes, and ion channel their score was increased by 1.
The defined process of selection and prioritization, which was a combination of automatic scoring and expert reviews by extensive target background analyses through telephone and live interactions, using the CarTarDis target database as the central shared knowledge platform, was further refined. Correlations between the experts and between expert and automatic scoring were evaluated, in order to make the selected workflow robust and applicable in future drug target selection projects. Overall we obtained good to excellent correlations between the systematic and the expert score. Pearson’s correlation coefficient was above 0.5 for all six categories. The best compliance could be observed for the category evaluating an involvement in lipid metabolism (r=0.81) but also for druggability and novelty, the observed correlations between expert and systematic scores were above 0.65. When comparing the expert scores for different categories among each other, OMICs was least correlated to the other categories. Absolute Pearson’s correlation coefficient was not exceeding 0.2 in the comparison with any other category (Supplementary Table 2). Furthermore, novelty was negatively correlated with all other categories. This mainly reflects that a high degree of novelty consequences a lack of information about a specific gene, as it is not yet fully explored and potential affirmative information is not yet available. In particular, druggability and novelty had a high, negative correlation coefficient (R=-0.58) which indicates that these two categories assess similar data in an opposed way.
This approach of parallel scoring ultimately resulted in 7 selected candidate targets in month 18 and target-specific validation work has been carried out. Between month 18 and 36, continuous screening for novel candidate targets yielded 15 new targets that were reviewed between month 37 and 48 using a similar procedure, in an attempt to identify targets with even better properties. Between month 37 and 48, further validation activities took place, as well as acquisition of additional omics data that was used as supporting evidence.
• To further characterize available, and develop new in vitro and in vivo mechanistic and efficacy models that enable the monitoring of the biological function of CVD targets.
To maximally support mechanistic validation of the selected CVD drug targets, we defined the best suited models and strategies aimed at endothelial cell function, atherosclerotic plaque development and lipid metabolism. We developed both in vivo (rodent) models and in vitro (cellular) models that allow evaluation of pharmacological and mechanistic aspects of the target validation projects.
Novel in vivo models are being generated aiming for:
i. Plaque rupture rodent model.
Quorics previously identified several genes whose expression was correlated to endothelial cell dysfunction and leakage. These might provide an opportunity to generate a plaque rupture model, which would be the first such model in the CVD field and highly desired to support drug development projects aiming at treating atherosclerosis. Inducible endothelial specific expression models for ETS2 and PLVAP and the conditional THSD1 knockout have resulted in transgenic lines. To limit the generation of mice and to save costs and time involved in the in vivo testing of the new mouse models, we first decided to test the constructs on functionality in vitro using bone marrow (BM) differentiated endothelial cells. These tests allowed for a preselection of expressing constructs and selection of a small number of transgenic lines for further analysis. From these lines, by breeding with a Cre delete line, we generated ubiquitous deletions for THSD1, and ubiquitous activations via lox-stop-lox deletions, of PLVAP and ETS2. Currently, we are testing endothelial permeability on these animals. The actual usefulness of these models will have to be established by in depth characterization and functional analysis, and will determine whether another partner will continue experimenting on them, after the official end date of CarTarDis. To successfully characterize these models, partners in this WP have decided to extend their collaboration for the coming year.
ii. Aging rodent model.
We decided to start creating a controlled ageing model based on the expression of progerin as a translational tool for research of vascular ageing and other age-related diseases. In order to achieve a mouse model with regulated expression of progerin cDNA, we used the elements that were established in tissue culture for mifepristone-controlled gene expression. As this system is built on the transient or stable co-expression of a regulator and a responder plasmid, we had to combine and test a one-vector construct. Such a setup is more convenient for generating transgenic animals, but it poses a challenge for the correct dosage and expression strengths of the two components, and as such, had therefore to be tested in tissue culture. We ultimately decided to attempt generating transgenic animals with two different constructs that have an either low or strongly driven regulator element. Both constructs were microinjected and yielded transgenic founders. Currently, we are in the process of breeding offspring for testing uninduced and induced gene expression of progerin, and hope to obtain positive results right after the official end date of CarTarDis. Given the encouraging outcome of the LPL experiments, the prospects are high that this model is functional.
iii) Target-specific rodent models.
Of the 7 selected targets, Of the 6 selected targets, only two justified the generation of in vivo rodent models to support target validation: LPL and IL6R. Unfortunately, genomic evolutionary analysis showed that IL6R and its downstream pathway was insufficiently conserved from man to mouse. To generate a model with controllable downregulation of LPL, we targeted mouse ES cells with a construct that was validated in a similar way as described above, and that contained a duplicated operator site for the mphR(A) regulator about 250bp downstream of exon 1. These cells were used to generate chimeric animals via blastocyst injections, which led to germ-line stabilized heterozygotes. In order to exclude an influence of the operator insertion alone on LPL expression, the mice were bred to homozygosity and analyzed for phenotypic changes at PG, and further characterized by measuring LPL expression at Umea. We observed an odd and surprising tail phenotype, but otherwise the mice are healthy and LPL expression is unaffected. These mice were bred with a number of strains that were generated at PG, and express a KRAB-based suppressing regulator in different configurations. Currently, 4 co-transgenic models are being bred at PG. For three of them (17 animals in total), we have generated first pilot data by applying mifepristone via i.p. injections or in drinking water. These models utilize co-transgenics with a heterozygous LPL locus, and therefore, we expected a limited effect given the strong post-translational levelling of gene expression. Expectedly, LPL expression was uninformative in most tissues, while expression in kidney, where post-translational correction is reportedly absent, we observed a clear correspondence and full suppression (to 50%, as expected in the presence of a wild-type LPL copy). We can therefore anticipate that mice with a homozygous LPL-targeted copy can show a full response to regulation in all relevant tissues.
When applicable, we have to opportunity to combine the new models with the ApoE3Leiden mouse, as drug-responsive model for hyperlipidemia and atherosclerosis, and PCSK9 AAV transductions to rapidly increase plasma cholesterol levels and susceptibility for atherosclerosis development.
Opportunities for functional assessment of some selected targets using therapeutic agents such as small molecules, antibodies and recombinant proteins were carefully considered for each target. Our model of choice is the well validated E3Leiden.CETP model at TNO. To further validate LPL as a target for atherosclerosis we decided to perform in vivo studies using an LPL activating tool compound (LP071). LP071 has previously been shown to ameliorate the postprandial response after an olive oil gavage by lowering triacylglycerol (TG) in hypertriglyceridemic apolipoprotein A-V-deficient mice and in a short term study in the APOE*3Leiden.huCETP (E3L.CETP) mouse model. An atherosclerosis study was designed for E3L.CETP together with pilot experiments to investigate dosing and administration routes (IP and oral). However, the three pilot studies using different dosages and ways of administration did not show any effects of treatment compared to vehicle controls. Compound plasma exposure levels were investigated and found to be in the low µM range with clear distinction between oral and IP delivery and high vs low dosing. Since previous studies had been performed using a different vehicle this was also investigated in a separate pilot dosing study without significant changes in plasma lipids. The results were followed by repetition of previous studies on WT mice. Mice were treated with IP injections for 4 days and subsequently challenged by an oral fat tolerance test. However, after analysing the results it was clear that activation of brown adipose tissue erased any differences between the treatment groups. The study was subsequently repeated in thermo-neutral conditions, ablating any effect relating to thermogenesis, where a minor effect on postprandial lipid clearance and LPL tissue activity could be found in the LP071 treated group. The limited efficacy combined with the difficulties to perform an atherosclerosis study in thermo-neutral conditions prompted a stop with further experiments using this tool compound. Currently these results are investigated to determine the reason of this irreproducibility. The subsequent atherosclerosis study is put on hold until we have obtained more information. To further validate the role of OSM in endothelial activation, we proceeded with an in vivo study testing both purified OSM protein and inhibitory OSM antibodies as therapeutic agents. First, a pilot study was performed to determine optimal dosage and a potential in vivo effect using markers of endothelial activation (inflammation) and plasma lipids as read-outs. Subsequently, a full atherosclerosis study will follow evaluating the effect of OSM/OSM inhibition on atherosclerosis development. To support the functional target assessment on atherosclerosis we established several in vivo imaging methods including optimized protocols for echocardiography (3D and full cardiac scan, Doppler function of coronary and carotid arteries) and PET/CT (using 18-FDG, ongoing oral lipid tracer 18-FTHA and FA analogue 11-C acetate) as well as optical projection tomography protocols for detection of protein factors in aorta and heart.
In parallel, several in vitro cellular models to study endothelial cell and macrophage function have been developed, optimized and applied in hypothesis-driven target validation studies, guided by expertise in WP4. Initially, the developments of cellular models were focused on endothelial cells, since there was considerable interest and experiences within the consortium related to this specific cell type in CVD. Proper function of the endothelial cells is crucial for vascular maintenance and homeostasis. Dysfunction of the endothelial layer plays an integral part in both the onset as well as progression of atherosclerosis. Furthermore, during target selection, endothelial function was integrated in the capabilities and pathways criteria. Therefore, many of the selected targets are expressed in endothelial and/or affect their function. New models focussing on other CVD relevant cell types (VSMC, monocytes, neutrophils) have been developed to answer specific questions. Below is a summary of major findings related to the use of cellular models for validation of some of the selected candidate CVD targets.
i) Endothelial cell models.
We evaluated the role of two genes that are used to generate the plaque rupture model (PLVAP and TNFAip8L1) on several aspects of endothelial (dys)function. PLVAP was show to promote endothelial survival and migration by indirect stimulation of the production of the pro-angiogenic factor S1P via eEF1A1. We observed that TNFaip8L1 inhibits endothelial cell apoptosis through the inhibition of caspase 8 activity and subsequent cleavage of caspase 3. These findings were being used to prepare two manuscripts that describe the endothelial function of the targets. We also assessed the expression of the prioritized targets and relevant pathway markers in healthy and stimulated primary endothelial cells using RT-qPCR. So far, mRNA expression could be confirmed for the targets MCL1, PECAM-1, LRP, LIPA, PDGFD, ANGPLT1, ABCA8, PPAP2B, EDNRA and IL6R. CYP17A1 and CTF1 showed no mRNA expression in endothelial cells.
Previously we found that the target OSM induces endothelial activation in Human Umbilical Vein Endothelial cells (HUVECs) and murine endothelial cells. We now discovered that that OSM also induces endothelial activation in cells from different vascular beds, Human Aortic Endothelial Cells (HAECs) and Human Microvascular Endothelial Cell (HMEC-1) as characterized by higher MCP-1, IL-6 and ICAM-1 mRNA levels, increased ICAM-1 membrane expression and increased MCP-1, IL-6 and E-selectin protein release. In addition, we observed a marked STAT1 phosphorylation, indicating involvement of the JAK/STAT pathway in OSM signalling. siRNA knockdown of the LIFR and the OSMR revealed that simultaneous knockdown is necessary to significantly reduce MCP-1 and IL-6 secretion and STAT1 and STAT3 phosphorylation after OSM stimulation. These data are reported in the manuscript: Inflammatory cytokine oncostatin M induces endothelial activation in macro- and microvascular endothelial cells and in APOE*3Leiden.CETP mice (submitted). PPAP2B mRNA is expressed in human endothelial cells and silencing of the gene resulted in enhanced endothelial activation in previous experiments (Figure 10A). As PPAP2B was deemed not druggable we focused on its downstream LPA receptors that are functionally closely linked and are druggable. Interestingly, of the six known receptors for LPA, the target of PPAP2B, LPAR6, was observed to be the highest expressed on endothelial cells. This suggested that higher levels of LPA, due to the knockdown of PPAP2B may enhance endothelial activation primarily through LPAR6. To further explore this hypothesis, we next stimulated the HUVECs with LPA, what led to increased expression of cytokines IL6 and MCP-1 as well as ICAM-1 and VCAM-1. These effects were aggravated when PPAP2B was silenced, again suggesting that lower LPP3/PPAP2B levels may cause endothelial activation through enhanced LPA signalling. To test the hypothesis that the effect of LPA would be mediated via LPAR6, shown to be the most prominent LPAR expressed in ECs in human plaques, we continued with LPAR6 silencing. Interestingly, silencing of LPAR6 in HUVECs inhibited LPA-induced endothelial activation, observed by the reduced expression of ICAM-1, VCAM-1, IL6 and MCP-1. However, simultaneous silencing of both LPAR6 and PPAP2B did not result in any additional cumulative effects on endothelial activation compared to those already observed after silencing PPAP2B alone. Based on these findings, we speculate that, although silencing of PPAP2B and LPA stimulation induce similar phenotypes in ECs, they may not share the same LPA-LPAR6 signalling. Our results also indicate a complexity and flexibility in signalling patterns among LPARs, that can be dictated by repression of PPAP2B in combination with some yet unknown factors in the atherosclerotic tissue environment, considering that silencing of LPAR6 did not have any further effect on the endothelial activation already observed after silencing of PPAP2B alone. These data are described in the manuscript: Integrated human evaluation of the lysophosphatidic acid pathway as a novel therapeutic target in atherosclerosis (submitted).
ii) Immune cell, smooth muscle cells and fibroblast cell models.
Since some questions arising from within the consortium cannot be answered by the use of endothelial cells, we have developed functional human monocyte/macrophage, smooth muscle cell and fibroblast models in order to provide new tools. Based on data from the Karolinska BIKE biobank lower levels of PPAP2B are associated with the larger presence of immune cells and higher levels of several LPArs in human atherosclerotic lesions. Therefore, we evaluated if stimulation with LPA and/or knockdown of PPAP2B had an impact on the expression of the different LPArs on monocyte and macrophages. Both LPA and PPAP2B knockdown didn’t affect the expression of the LPARs on these cells in vitro. These data are incorporated in the manuscript on PPAP2B validation (noted above). To elucidate the trans-signalling effect of IL6r we stimulated monocytes, macrophages and fibroblasts with or without siRNA mediated knockdown of the IL6r in these cells. No clear effects were seen, which we believe are due to the high levels of IL6 that are produced by these cells. New experiments have been scheduled that take this into account.
iii) Foam cell models.
Biobank expression data from WP3 showed that many of our targets (OSM, PPAP2B, IL5 and CECR1 co-localize with foam cells. We ran experiments to turn monocytes/macrophages into foam cells, to study the role of these targets in these cells.
iv) Peripheral Blood Mononuclear Cells (PBMCs).
We also performed experiments using freshly isolated human PBMCs to elucidate the effect of CECR1 on immune cells. According to literature CECR1 should affect the immune status, by inducing the differentiation of immune cells towards an immune suppressive phenotype and subsequent induction of proliferation. Using flow cytometry, we could not identify an effect of CECR1 on B-cells and T-cells. Interestingly, in a pilot experiment we found an inverse relation between CECR1 levels and the number and activation of neutrophils, suggesting that neutrophils may be the real target of CECR1. Currently we are expanding the number of patients to substantiate this claim.
v) Neutrophils.
Recently the target CECR1 has been implicated to play a role in the activation of neutrophils. These changes in neutrophil biology are accompanied by compromised endothelial integrity, endothelial cellular activation, and enhanced inflammation. Together, this suggests that CECR1 may play role in atherosclerosis development by regulating neutrophil activation. To assess the effect of our target CECR1 on neutrophils in vitro, freshly harvested human neutrophils were stimulated with adenosine. Adenosine strongly induced neutrophil activation as seen by increased release of the atherosclerosis promoting factors MPO, MMP9, and neutrophil extracellular traps (NETs). Co-stimulation with CECR1 resulted in a reduction of MPO, MMP9, and NETs. These results suggests that CECR1 inhibits neutrophil activation by the conversion of adenosine to inosine. The subsequent effect of CECR1 inhibited activation on other cells and the process of atherosclerosis needs further investigation, which will continue post CarTarDis.
vi) Mural cell models (VSMCs).
To elucidate the seemingly complex role of IL6-IL6r signaling, we performed a series of in vitro experiments in which we silenced either IL6 and/or IL6r in various cell types. Thus far we observed no differences that suggest any differences in IL6 signaling. Since the production of IL6 in cultured cells is still relatively high even under knockdown condition, we decided to also include a neutralizing antibody, to really block the signaling cascade. These experiments are currently being performed and will continue post CarTarDis.
• To develop and optimize cutting edge molecular assays to accurately monitor proteins, RNA and metabolites in preclinical and clinical biosamples relevant to CVD.
- Biobanks:
To support the validation of candidate targets in human cardiovascular disease, we had access to two human biobanks containing well-annotated, appropriately stored, high quality tissue biosamples with high relevance to CVD. Socrates (LUMC) is based on material obtained during organ transplantation (coronary and aorta wall samples), and BiKE (KI) is focussed on surgical specimens, in particular carotid endarterectomies. Socrates (LUMC) is largely based on perirenal aortic patches (from over 500 donors) that are collected during kidney transplant. Unlike other atherosclerotic biobanks, which are all based on end-stage surgical specimens or post mortem samples, material in Socrates is from relatively healthy individuals who fulfilled the Eurotranplant criteria for kidney donation. As such material in Socrates covers the full life-span (5-80 years of age), as well as the full spectrum of atherosclerotic disease. Removal of the aortic patch along with the renal artery is a standard procedure in order to preserve the integrity of the renal artery of the donor graft. At the time of transplantation the aortic patch is removed, split in half with one half frozen and the other half formalin fixed. A Movat staining is performed for each section and the stage of atherosclerosis graded (adapted AHA classification as proposed by Virmani). Material from the Socrates biobank was exchanged between partners of WP3 and used extensively to characterise expression of the selected targets and associated pathway biomarkers in the tissue samples representing various phases of CVD. The BiKE resource (KI) contains biosamples from patients who are scheduled for surgery for symptomatic or asymptomatic carotid stenosis and are included prospectively after written informed consent. After surgical retrieval, the endarterectomy/plaque is rinsed in physiological buffer and divided transversely at the culprit of the lesion. The proximal part is snap frozen in liquid nitrogen and the distal part fixed in formaldehyde (or embedded in O.C.T. compound and frozen). Blood samples are obtained simultaneously for plasma preparation and isolation of PBMCs for genotyping and gene expression analysis. Plaque tissue is utilized for isolation and preparation of mRNA, miRNA and cDNA. Furthermore through the highly optimized workflow, BiKE allows a rapid and efficient identification and validation of potential target candidates, incorporating transcript profiles, in situ protein detection and correlations with clinical data, genotypes, plasma levels and morphology. Both LUMC and Karolinska continued to update their human biobanks. Sampling procedures were adapted in order to collect samples suitable for further downstream analyses, which include immunohistochemistry, mRNA and microRNA detection, and lipid mass spectrometry. Finally, a preclinical biobank was set up at TNO that mirrors the human biobanks and enabled cross-species molecular analysis of the selected CVD targets and their pathways.
- Sample and data exchange:
The consortium partners developed a multi-partner sample exchange program whereby the same CVD tissue sample was analysed by multiple partners with their specific molecular staining methodologies. Protocols for analysis of plaque samples were optimized by the different partners and in parallel, best practice knowledge on staining protocols was exchanged. This workflow was used for the prioritized CVD targets’ studies. Snap frozen human carotid plaques (from KI) were sent to Imabiotech for lipid and metabolite imaging by Mass Spectrometry Imaging (MSI). Sections from the same plaques were sent back to KI for Immunohistochemistry (IHC) imaging and to Bioneer for mRNA In situ hybridisation (ISH) imaging. MSI, IHC and ISH pictures were shared between partners. Since routine tissue processing techniques interfere with MS-imaging application Imabiotech relies on dedicated native tissue samples for its mass spectrometry imaging. As this limits the application of MSI, Imabiotech has investigated methods to use formalin-fixed tissue for MS-imaging. To enable data sharing among CarTarDis consortium partners, Morphisto has established a virtual microscopy environment that allows on line data sharing. Imabiotech developed a server-based data storage and exchange facility that allows easy access with partners. LUMC developed software to share IHC images obtained using their Philips digital microscopy system operative.
- Innovation in Molecular staining methodologies:
The partners in the CarTarDis consortium WP3 further expanded their expertise in molecular staining methodologies. For staining of mRNA and miRNA, an automated in situ hybridization (ISH) platform was established by Bioneer using branched DNA probe technology (RNAscope) halfway the project. Also, Bioneer developed a double staining method based on fluorescence to visualize mRNAs and cellular markers in the same atherosclerotic tissue, e.g. to document the expression of PPAP2B and LPAR6 in luminal endothelial CD34+ cells in the plaque. Using these methods, Bioneer completed the optimized screening and evaluation of the expression of mRNA from the first selection of candidate target genes: OSM, OSMR, LIFR, CECR1, IL5, IL6R, PPAP2B/LPP3, LPAR1, LPAR2, LPAR5 and LPAR6 in eight FFPE-blocks from the Socrates biobank representing the following stages: AIT, IX, PIT, EFA, and LFA of atherosclerosis development. The complete screening was performed using the brown-staining protocol using DAB, and a subset of the FFPE-blocks representing stages IX, EFA and LFA were subsequently stained using the red dye FastRed. The use of the red dye, enables combining the mRNA-ISH signal with immunofluorescent staining of cells to provide evidence of the origin of the cell expressing the genes.
For staining of proteins, Morphisto optimized immunohistochemistry (IHC) methods for the selected targets and relevant pathway protein biomarkers and receptors, in good alignment with the IHC development at Karolinska and LUMC. In addition, Morphisto optimised classical histochemical staining methods and improved throughput of their staining workflow. Staining tasks for the respective targets was divided among the partners, whereas in addition, Morphisto executed expression profiling of the top 20 targets in human lesions from the Socrates biobank (LUMC) that spanned the full cardiovascular disease spectrum. Double and multiple IHC protocols were developed by KI and LUMC to investigate and validate the cell type specific expression and localization of the different targets. LUMC developed and validated a versatile protocol that allows for sequential staining cycles on a single sample, thereby allowing simultaneous appreciation of multiple markers (IHC analysis of 7 protein biomarkers was demonstrated).
For staining of small molecules including lipids, Imabiotech applied their mass spectrometry imaging (MSI) platform to semi-quantitate specific molecules such as cholesterol esters (CE), Lysophosphatidic acid (LPA), phosphatidylcholine (PC), phosphatidic acid (PA) and others. These were particularly used to establish co-localization of selected lipid substrates and aspects of the PPAP-2B pathway. In addition, Imabiotech developed specific assays to measure and localise tissue levels of adenosin/inosin in atherosclerotic samples from the Socrates biobank that were linked to expression levels of CECR1 mRNA and protein. To further innovate their MSI capabilities, Imabiotech developed methods to use formalin fixed tissue for MS imaging, thus expanding the application of their MSI platform to evaluation of non-surgical specimens.
For analysis of gene expression in specific tissue regions, KI has optimized laser capture microdissection (LCM) microscopy for isolation of specific human carotid plaque regions (fibrous cap, shoulder region and necrotic core). LCM is a technology that allows isolation of single cell populations under direct microscopic visualization, followed by various downstream applications as RNA transcript profiling, cDNA library generation, proteomic studies and discovery of novel signalling pathways. By using LCM, we have successfully isolated α-actin positive cells (αSMA+) in the fibrous cap of human endarterectomies. RNA quality and concentration were evaluated after every LCM by bioanalyzer and RNA integrity number (RIN) was estimated to be between 3 and 6. RNA concentration was approximately 0.3-3 ng/μl. We demonstrated that analysis of expression of housekeeping genes (GAPDH, RPLP0) by quantitative RT-PCR (Taqman) showed a RNA concentration dependent curve, indicating the LCM-gene expression assay is ready for analysis of specific target genes.
- Application to validation of selected targets:
We successfully applied complementary techniques to study the prioritized CVD targets identified in WP1 and WP4. A combination of metabolic data from WP1 (HMGU) and genetic analysis (Icelandic Heart Association, IHA) suggested an association between lysophosphatidic acid species and incident of plaque development. Previously, using samples from the BiKE biobank, ImaBiotech showed an accumulation of lysophospholipids in the shoulder of the plaque. A combination of lipid imaging together with specific IHC (from KI) identified the overlay regions in the plaques between lysophospholipids and lipid phosphate phosphohydrolase 3 (LPP3/PPAP2B). Along similar lines Karolinska, LUMC, Bioneer and Morphisto exchanged samples in order to validate the prioritised targets by ISH and IHC. The successful integration of complementary techniques to study the prioritized CVD targets is illustrated by the work on PPAP2B, the lysophosphatidic acid (LPA) pathway and atherosclerosis susceptibility. Based on the genetic data from IHA, the PPAP2B gene encoding for lipid phosphate phosphohydrolase 3 (LPP3) was prioritized as one of the primary targets in the project. Using samples from the BiKE biobank, ImaBiotech showed an accumulation of lysophospholipids in the shoulder of the plaque. A combination of lipid imaging together with specific IHC (from KI) identified the overlay regions in the plaques between lysophospholipids and lipid phosphate phosphohydrolase 3 (LPP3/PPAP2B). A similar combination of techniques was applied to detect LPAR6 protein by immunohistochemistry staining (KI) and LPAR6 mRNA by in situ hybridization (Bioneer) in human atherosclerotic tissue. Data generated are intrinsic part of all of the target validation projects and of the corresponding manuscripts that have been prepared.
During the last months of the CarTarDis project, Bioneer and Morphisto undertook the task of screening the entire set of top 20 target genes defined in WP1 and WP4 on mRNA and protein level respectively. The samples were provided by the Socrates biobank. After initial screening of the extra blocks a set of nine blocks were selected for the mRNA-ISH screening of the 20 targets. At Bioneer, 20 genes were screened using the automated staining platform and the RNAscope branched-DNA probe technology with red staining. All slides were scanned and evaluated for mRNA expression in different compartments (adventitia, media or intima) of the different stages of atherosclerotic plaques. Additionally, ISH-screening of OSM, OSMR and LIFR on the same nine samples mentioned above was performed to support the studies and publications on OSM. Together with KI, Quorics, LUMC and ImaBiotech a multimodality imaging and staining study was also established using a cryo-preserved carotid plaque for BiKe. CECR1-ISH, parallel CECR-IHC and MSI to image the presence of inosine and adenosine was performed on serial sections prepared and distributed by ImaBiotech for multimodal imaging. Similarly, a multimodality imaging and staining study on another set of serial sections, involving CD31/PECAM-ISH, PLA2-IHC and MSI to image LPA and LPC phospholipid species was also done to support the technology focused paper and poster to be presented at the AHA-2017 meeting.
• To apply molecular assays to further characterize biosamples of two comprehensive well-annotated human CVD biobanks to investigate and validate the contribution of specific pathways to CVD.
Analysis of the two unique human biobanks in the CarTarDis consortium (Socrates, Bike) that contain precious biobanked human material from CVD patients in all stages of disease has provided insightful data to support target validation including co-expression of the selected targets and associated pathways and evidence for their clinical relevance. The section above outlines how the novel molecular staining protocols were applied to human CVD biosamples from the Socrates and Bike biobanks in the development and validation of the methods and of the evaluation of the preselected candidate drug targets. In the final year of the CarTarDis project, the implemented and developed methods were used to visualise the top 20 targets selected in our workflow using samples from the Bike and Socrates biobanks.
• To develop and exploit this newly formed pan-European modular pharmaceutical target discovery network in cardiovascular disease.
The intense collaboration between the CarTarDis consortium partners during the development of novel tools and models and during the validation studies of the selected targets provided a network among the multidisciplinary experts as originally planned. This has yielded several spin-off collaborations between some partners, whereby the unique expertise of a partner contributed to contract research projects of other partners (details not to be disclosed due to confidentiality) or whereby partners involve each other in related activities (eg close interaction between biomarker networks of Netherlands (prof A van Gool) and Denmark (dr K. Holmstrøm). In addition, the joint consortium meeting with the three FP7 CVD consortia widened the window of opportunity for collaboration with other peer scientists. This yielded several bilateral collaborations in contract research, joint publications of opinion papers and submission of novel joint research proposals. The topics of these novel interactions include cardiovascular research, translational biomarker development and application of molecular technologies in translational research.
Potential Impact:
Guided by the nature of the EU FP7 call (SME targeted ambitions) we defined the overall objectives of the CarTarDis project as: 1. Identify and validate novel CVD drug targets, 2. Develop in vitro and in vivo tools to support CVD drug development, 3. Increase the knowledge and competitive position of the participating partners (notably the SME’s) in drug discovery. The potential impact is best to be described for these three categories.
i) Novel CVD drug targets.
Directed by the experienced pharmaceutical scientists in our team and advised by experienced drug developers and entrepreneurs in our External Advisory Board, we designed a rational approach to accurately reflect the pharmaceutical process, adopting best practices from pharmaceutical industry from its start. The selection filters were inspired by the 5R strategy of AstraZeneca and included Genetic/omic evidence, Druggability, Novelty, Feasibility and Pathways. A database and target selection workflow was designed that allowed us to select and prioritise candidate CVD drug targets. For each preselected target, a target champion was appointed who led a multidisciplinary team of consortium scientists to define the key experiments to generate data to support the validation of that particular target. All teams were guided and supported by the experienced pharmaceutical drug hunter scientists and clinical cardiovascular experts in our consortium, together forming strong teams to reach the objective of identifying novel validated cardiovascular drug targets. First, available and emerging molecular data from three large independent cardiovascular cohorts was mined for statistical association with cardiovascular disease phenotypes, with a preference for genetic associations. Then, validation of the candidate targets was done by two parallel approaches, in which mechanistic validation was performed using novel in vitro (cellular) and in vivo (rodent) preclinical models in which the candidate targets were studied regarding mechanism of action and role in particular aspects of the cardiovascular processes and clinical validation was based on two unique human cardiovascular tissue biobanks and several high resolution molecular analysis methods using which the spatial expression of the target and target-related pathways in cardiovascular tissue samples was studied on RNA, protein and small molecule level. Of the 381 preselected candidate targets, we prioritised 7 targets, i.e. PPAP2B, OSM, CECR1, LPL, IL6R, IL5, ApoBD. Seven multidisciplinary target validation teams were formed involving all partners of the CarTarDis consortium. In line with current practice in pharmaceutical drug discovery, we experienced the complexity of human biology and the new data supported further validation for only 3 of the selected targets (PPAP2B, OSM, CECR1). For some, we identified downstream proteins as potentially better drug targets than the originally preselected target and focussed our validation experiments on those (for instance LPAR6 was a more druggable target whereas the initial hit was PPAP2B. We sought interaction with parties for identification of potential adopters of the validated drug targets from the CarTarDis project for subsequent drug development. These included (semi-)industrial partners in our network, notably AZ and the European Lead Factory, and our External Advisory Board, notably Menzo Havinga and Erik Lund. This did not lead to start of an adoption process before the close of the CarTarDis project. Although we progressed significantly, the validation status of the selected targets was not at a level yet to merit subsequent investments by private partners. We finished the validation projects with clear view on additional experiments that are needed beyond the project to provide ultimate support for the chosen targets. All studies will be published to impact the scientific cardiovascular field, using peer-reviewed publications and posters at CVD conferences (such as the recent American Heart Association 2017).
ii) in vitro and in vivo tools to support CVD drug development.
The partners in the CarTarDis consortium developed novel innovative tools that support the validation of candidate drug targets in cardiovascular disease and other therapeutic areas, and support the development of drugs targeting those novel targets. Among the ICT tools that were developed, the interactive database that we used to prioritise the preselected CVD drug targets (developed by HMGU) received much appreciation from our External Advisory Board, outlining that such approach was rarely been followed in the academic communities and would be worthwhile disseminating (manuscript is in preparation). The data exchange solutions developed by consortium partners in (notably LUMC, Morphisto, Bioneer, Imabiotech) will facilitate future interactions of these partners in bilateral and multi-partner collaborations. The in vitro (cellular) models and in vivo (rodent) models that were developed (detailed above) will impact the cardiovascular research field as they provide new capabilities to translational research, either using the developed assays or methodologies. To illustrate, the novel targeting strategy using to generate a tissue-specific expression modulation of LPL (developed by Polygene) will be applicable to many more rodent models in the future (patent filed). The functional cardiovascular cellular models in combination with siRNA mediated knock-down of gene expression used to study the mode of action of specific genes, will not only impact the cardiovascular researchers but also provide a template workflow that can be used by other researchers (multiple publications published or in preparation). The molecular staining methodologies are described in explicit detail above. Notably, the mass spectrometry imaging technology to highly specific visualize particular small molecules in tissue slides (Imabiotech) was rather unknown in the cardiovascular field, whereas we could clearly demonstrate the added value of this approach, particularly in combination with other imaging modalities (multiple posters and papers have been prepared). The branched DNA approach to visualize specific mRNA transcripts in tissues in an amplified manner (developed by Bioneer) has the potential to significantly impact molecular analysis of tissues by increasing the sensitivity of spatial gene expression analysis.
iii) Competitive position of the participating partners (notably the SME’s) in drug discovery.
The SME’s in the CarTarDis consortium participated strongly in these two validation activities by contributing their unique technological expertise as part of the validation teams. Throughout the project, we paid particularly attention to the educational aspect of the pharmaceutical process by outlining the rationale behind particular decisions in the pharmaceutical pipeline and involving all partners in this process. Judged by the feedback of the partners (particular the SME’s) during the project and upon finishing the final period, this was much appreciated. Particularly the knowledge of participating partners with still limited experience in the pharmaceutical process was boosted to an experienced level that impact their ability to interact with pharmaceutical researchers in the future. The joint consortium meeting with the three FP7 CVD consortia (Sept 2016) increased the exposure of the partners and widened the window of opportunity for collaboration with other peer scientists. This yielded several bilateral collaborations in contract research, joint publications of opinion papers and submission of novel joint research proposals. The topics of these novel interactions include cardiovascular research, translational biomarker development and application of molecular technologies in translational research. As such, the CarTarDis consortium members have become more attractive collaboration partners in bilateral and multi-partner collaborations, one of the key objectives of the CarTarDis project.
List of Websites:
http://cartardis.eu