Final Report Summary - SYSCLAD (Systems prediction of Chronic Lung Allograft Dysfunction)
Executive Summary:
Recognized as the next major step forward in medical research, Systems Medicine (SM) offers a groundbreaking approach to understand and unravel complex diseases mechanisms and potentially overcome many of the current limitations of drug discovery. SM is also expected to lay the foundation of predictive, personalized, preventive and participatory (P4-Medicine), changing the way we practice medicine from a reactive to a proactive mode and focusing more on wellness rather than on disease. The FP7-funded research project SysCLAD, coordinated by Prof. L. Nicod, is one of the flagship initiatives of Systems Medicine in Europe, bringing together leading clinical centres in lung transplantation (LT), academic groups and industrials to i) identify and validate a Clinical and Molecular Signature Predictive of Chronic Allograft Lung Dysfunction (CLAD) and ii) develop a personalised medicine tool, based on mathematical representations of disease mechanisms, that will allow a prediction of CLAD resulting in Early and Specific Interventions LT patients at high risks to develop CLAD. The large knowledge base of integrated clinical/omics data produced for the validation of the SysCLAD model will be an invaluable resource for further SM-driven research in the context of other chronic respiratory diseases of major incidence such as Severe asthma, Chronic Obstructive Pulmonary Disease or Interstitial Pulmonary Fibrosis
The objectives of WP1 were: to obtain authorizations and ethical clearance from regulatory authorities from France, Belgium and Switzerland to merge COLT and STCS cohorts in order to recruit at least 840 candidates for lung transplantation and follow-up at least 400 LTR for 3 years; to define and disseminate Standard operating procedures (SOPs); to store and deliver samples to other WP; to guarantee the quality of data generated in the project. All authorizations and ethical clearance were obtained, with some significant delay for CNIL clearance on the environmental exposure study. All SOPs were disseminated to all participating centres. The objective of 840 candidates for lung transplantation has been exceeded and 422 LTR have now been follow-up during 3 years. Clinical and functional data from these 422 patients were reviewed and harmonised by successive adjudication committees in order to stratify patients into the sub phenotypes of CLAD as Stable, BOS, RAS, mixed or other. High quality biological samples and data from these patients at M6 and M12 post-transplantation were delivered to partners for omics, microbiome and immunological analysis to start the planned analysis. The only deviation from the original work plan was linked to the environmental exposure study which was started later than planned due to delayed authorisations. However, all work was carried out by UJF to correct this delay and UJF provided accurate estimates of the chronic exposure to ambient air pollutant exposure at the individual level and assessed the association between chronic air pollutant exposure and FEV1 level and decline.
In WP2, Samples from the SysCLAD biobank (blood, BAL) were analysed by a wide range of analytical methods covering transcriptomics (mRNA, miRNA), full exome sequencing, proteomics (SELDI-TOF, iTRAQ LC-MS/MS), immunological bioassays and microbiome characterisation according to the various CLAD sub-phenotypes identified in WP1. All data were fed into the global database generated by BIOMAX before complete integration alongside the clinical parameters into the numerical model. This WP led to the identification of signatures of each CLAD sub-phenotypes according to differences in genes and proteins expression in lung transplant recipients, and allowed the identification of specific patterns of pulmonary microbiota composition varying in concert with the underlying alveolar inflammatory background. In WP3, a stable SysCLAD Information System was created, where clinical, experimental and data from external databases are integrated and linked. A dictionary-based literature analysis tool with graphical integration of the results adapted to the needs of the clinician community was developed and disseminated.
In WP4, the computational model developed by Novadiscovery in the context of the SysCLAD project comprises three fundamental bricks: i) a disease model, which describes, computationally, the pathogenesis of CLAD; ii) a clinical model, which translates the outcome of the disease model into a real clinical variable, such as those used to assess the pulmonary function of a patient; iii) a virtual population, which describes the inter-individual variability that characterizes the natural history of CLAD.
In summary, the main objective for this work package was not fully achieved. However:
• The mathematical model is not far from being completed.
• Instead, we developed a logic-based model (it was not initially planned) that enabled to achieve some of the objectives.
• We developed a software application to simulate any combination of perturbations (pollutants, infections, GERD, etc…)
• We developed a Virtual Population with unrestricted values for patients’ descriptors.
The framework established in WP5 ensured efficient exchange of scientific information both within the consortium and to the wider scientific and industrial communities through a vast number of dissemination activities to all stakeholders via the numerous communication channels available to the partners and establishing new ones. In addition, it allowed an efficient follow-up and management of ethical, regulatory and societal issues that arised during the project.
All procedures for internal communication, technical and financial reporting/IP monitoring at WP level, partner level and consortium wide were efficiently implemented. SysCLAD governance was extremely efficient to rapidly take and act upon consortium decisions. Procedures were specifically developed to be adapted to the nature of the SysCLAD project (i.e. SME-centric) allowing for a thorough yet agile management and follow up of all activities of the project. Monthly TCs, led by FINOVATIS were implemented to coordinate technical, financial and administrative activities and identify any potential issues and define corrective actions in a rapid manner. No significant deviations from the original work plan were notable from a management standpoint.
Project Context and Objectives:
With more than 32,000 procedures performed around the world in the least 30 years, lung transplantation (LT) has become the standard of care for selected patients with advanced lung diseases. Unfortunately, LT procedures are often at risk of chronic lung allograft dysfunction (CLAD), of which the two most common pathologic presentations are i) a remodelling of small airways resulting in an obstructive ventilatory pattern “Bronchiolitis Obliterans Syndrome” (BOS) and ii) a restrictive ventilatory pattern “Restrictive Allograft Syndrome” (RAS). Almost 50% of LT recipients will develop CLAD within 5 years post LT, rising to 75% after 10 years. The median survival figures from the time of diagnosis remain very poor for both early-onset CLAD (1.5 years) and late-onset CLAD (2.5 years). In clinical routine, the decline of the forced expiratory volume in one second (FEV1) is used as a first indication of CLAD onset once differential diagnoses (e.g. acute rejection, active infection and/or proximal bronchial complications) have been ruled out. Treating CLAD remains a challenge for several reasons: the main pathways, molecular events and triggers leading to Epithelial-Mesenchymal Transition (EMT) within the small airways or transformation of mesenchymal stromal cells, are not well understood and there has been, to date, no real attempts, due to the lack of tools, to integrate risks factors for a specific lung recipient to anticipate lung function decline before irreversible lung function loss. In the absence of robust guidelines and compounds active on EMT, CLAD treatment is often initiated late, chaotic and involves an increase in immunosuppressant which often leads to associated opportunistic infections, kidney failure, re-transplantation and, ultimately, death. An early prediction of CLAD based on principles of systems biology and personalized medicine would represent a major breakthrough and open a new area marked by early interventions based on risk stratification and personalized pathways modulation. Early recognition and specific intervention in patients at risk of CLAD would undoubtedly improve outcomes.
The major aim of this project is to build the first SysCLAD computational model to be used by clinicians and researchers. The model will identify and validate a “signature” of CLAD integrating multilevel data in order to i) identify and mitigate the risk of chronic allograft dysfunction and ii) allow for early intervention before irreversible damage. The SysCLAD computational model created in this project should allow prediction within the first year post LT of which recipients are at risk to develop BOS or RAS by 3 years post LT. This prediction, based on a mathematical and computer model developed through a systems biology approach, will be derived from the complete integration of large experimental data (clinicome, environmental data, omics, microbiome, and immunological assays) collected from both donors and recipients (See Annex - Supporting Fig. 1.1). The project will establish a prospective cohort of LT candidates (European Cohort Of Lung Transplantation (ECOLT)), which will expand the ongoing Cohort Of Lung Transplantation (COLT) (http://www.clinicaltrial.gov/ClinicalTrials.gov Identifier: NCT00980967) recruited since June 2009.
The SysCLAD model is constructed using data from a first cohort of 200 LT recipients and refined/validated using the new set of 200 LT data generated by ECOLT. In order to develop the SysCLAD model, the following five objectives are pursued in an integrated workflow:
1. Recruitment of LT candidates and LT recipients (ECOLT) was initiated in November 2012: 500 LTR were already included between June 2009 and January 2012. In addition to this, the inclusion of 200 new candidates (surviving more than 3 months post LT) resulted in availability of more than 700 LTR by January 2013 at project start. Of these, 400 LTR will reach 3 years follow up by July 2014 (M18) and be used to build, calibrate and validate the SysCLAD prediction model. In parallel, standard operating procedures related to sample collection (timing, type and mode of data collection) that have already been established will be extended to others members of the consortium for global data harmonisation.
2. Proteomics, transcriptomics and immunological assays of minimally invasive sampled fluids (blood and bronchoalveolar lavage - BAL) will be carried out and used to achieve objective 3.
3. A comprehensive clinical research information system (CRIS) centralising all the clinical and experimental data (environment, phenotype, microbiome, biology, omics) collected from donors, recipients in a longitudinal fashion will be built.
4. CLAD signature profiling. The data generated from objective 3, together with additional data from the literature will be integrated and modelled in order to build the first signature of CLAD.
5. Validation of SysCLAD. The SysCLAD model will be refined and validated by iteration in consecutive sets (see figure below).
To yield substantial benefits to drug discovery and development efforts, disease modelling needs to embrace the full spectrum of mechanisms and their interactions, from genes to populations. Our approach, which purports to design multi-scale mathematical disease models plugged in virtual populations, represents a powerful link to bridge systems biology and clinical applications together for the ultimate benefit of clinicians and patients.
The overarching objective of SysCLAD is to deliver personalised medicine and drug comparative-effectiveness prediction capabilities over a specifically characterised target population. The systems biology-driven modelling approach includes all the biological components that are thought to play a role in the course of a disease process, up to the clinical events, by plugging the disease model into the virtual population. The virtual population represents the population of interest. Each virtual patient of the population is characterised by a vector of descriptors that translate biological and environmental parameters involved in the course of the disease. Applied to CLAD modelling, the virtual population will be created from environmental, clinical and molecular vectors.
The specific objectives of each scientific work package were as follows:
WP 1: SOPs, ECOLT COHORT DESIGN, BIOBANKING (CHUN)
• To edict and set up Standard Operating Procedures (SOPs) adapted to Lung Transplantation in the 13 participating Centres of Lung Transplantation.
• To recruit 840 candidates for lung transplantation from whom 700 LTR
• To follow-up 400 Lung Transplant recipients for 3 years
• To constitute a database of clinical and functional data from 400 LTR
• To constitute a certificated biobank from 400 Lung Transplant recipients
• To deliver data and samples to investigators from WP 2 -5 as 2 successive cohorts of 200 patients with a 3 years follow-up
• To characterise the environmental exposure of LTR
WP 2: GENERATION OF BIOMARKERS TO BUILD SYSCLAD (UJF)
This WP was specifically dedicated to the generation of experimental data for the construction and the validation of the SysCLAD model. It is constituted of several parts including blood transcriptomics, bronchoalveolar lavage (BAL) and blood proteomics, recipient polymorphism analysis on blood, analysis of LTR lung microbiomes, and immunobiological assays on blood and BAL. The results of this WP allow the definition of transcriptomic, proteomic and immunobiological profiles of CLAD that can be used to build the SysCLAD model. In addition, these results will lead to the identification of new biomarkers for CLAD prediction and allow a better understanding of CLAD disease mechanisms.
WP 3: ESTABLISHMENT OF THE SYSCLAD KNOWLEDGE MANAGEMENT SYSTEM (BIOMAX)
Work package 3 will establish the content and technical infrastructure necessary to connect the clinical and experimental data generated within WP1 and 2 with the SysCLAD model developed in WP4. Specific objectives are:
• To provide a secure, federated software environment to integrate existing public and project-specific data sources and establish the data flow between project partners
• To semantically map the different SysCLAD clinical and experimental data types, information from public repositories and knowledge extracted from the literature
WP 3 will generate the first specific information network around CLAD, enabling WP4 to mine the molecular mechanisms involved in the process of developing BOS/EMT. It will bring together current general knowledge about human molecular process from existing resources such as KEGG with CLAD-specific, but unstructured knowledge available in the literature and available clinical and experimental data. This information network integrating prior knowledge with disease-specific data will be mined for connections and entities currently not known to participate in CLAD or involved in new roles. Technologically, WP3 will create the first CLAD research and therapy-specific graphical user interface supporting clinical workflows and decision processes. Technical integration between currently disjointed CRIS, literature based biomedical knowledge, molecular data and mathematical model will enable clinical researchers to apply the system medicine workflow. An existing literature mining process will be adapted to CLAD-specific information e.g. through development of a dictionary of CLAD-related terms for literature mining. Finally an existing knowledge management technology will be adapted to new experimental technologies and corresponding data types and formats.
WP 4: SYSCLAD NUMERICAL DESIGN (ND)
The primary WP4 objective was to develop a fully operational numerical model of CLAD to enable patient-level prediction of CLAD occurrence. At the beginning of the project, it was thought that the global model would consist in 6 sub-models:, i) Non-allo-dependent inflammation (prolonged ischemia, ischemia-reperfusion lesion, lymphatic drain disruption, denervation, gastro oesophagus reflux, etc.); ii) Allo and autoimmune mechanisms; iii) Infections (viral: CMV, rhinovirus, etc.; bacterial); iv) Innervation (without the structural and organisational aspects) with mediators, stimuli and targets linking to the nervous system; v) Model of the bronchiolitis epithelium; vi) Functional anatomy of bronchioles (from a phenomenological standpoint): stenosis growth and its impact on airway flow, which would yield the clinical criteria to stratify patients (a quantitative criterion translated into a binary “>0 or <0” variable).
The main scientific innovation from this WP was expected to be an overall representation of all mechanisms involved in CLAD disease progression. In particular, new knowledge was expected to emerge in the form of new prognosis biomarkers and new treatment efficacy biomarkers. Additional expected innovations were the structured accumulation & evaluation of knowledge available on the CLAD disease in the scientific literature through the development of a novel knowledge base and a graphical & comprehensive representation of the disease’s sub-mechanisms. The main expected technological innovations were: a CLAD risk prediction algorithm, which should take into account patient-specific descriptors. In particular, the development of a bedside personalised medicine web application was planned, for clinicians to (i) predict the likelihood of CLAD occurrence and (ii) predict optimal treatment responses. On the longer term, the CLAD numerical model associated with its virtual population should enable new potential therapeutic targets to be identified for new therapy development.
Project Results:
Significant results and achievements from each work package of SysCLAD
WP 1: SOPS, ECOLT COHORT DESIGN, BIOBANKING (CHUN) PARTNERS INVOLVED: CHUV, UJF, UZH
1. Implementation of Standard Operating Procedures (SOPs) adapted to Lung Transplantation in the 14 participating Centres of Lung Transplantation.
2. Recruitment of 1608 candidates for lung transplantation of which 1107 are LTR
3. Follow-up 422 LTR for 3 years
4. Development of a clinical and functional harmonised database for 422 LTR, with Biomax partner, freely accessible to all partners
5. Biobanking Certification obtained for samples storage
6. high quality biological samples and associated data collected from the well characterized 220 SysCLAD LTR patients have been delivered to investigators from WP 2 -5
7. Accurate estimates of chronic exposure to ambient air pollutant exposure at the individual level
8. First study suggesting an association between chronic exposure to ambient air pollutants and FEV1 level in LT patients
WP 2: GENERATION OF BIOMARKERS TO BUILD SYSCLAD (UJF) PARTNERS INVOLVED: CHUV, CHUN, GATC
Exome analysis (GATC)
1. Evaluation of all exon sequencing kits. Agilent’s SureSelect All Exon V5 have been used for the analysis of exome sequencing samples provided by SysClad.
2. Both Illumina’s Sequencing Systems HiSeq 2000 and MiSeq were suitable and were utilized for sequencing all exon samples.
3. Several possibilities to do comparative analysis of SNPs and InDels were evaluated to identify the most appropriate approach. GATC has decided to use a third party web-based analysis tool (see also “deviations”).
4. All samples were successfully sequenced and analysed.
Transcriptomic mRNA and miRNA Analysis (CHUN)
5. Extraction of RNA and RNA quality verification (>80% passed)
6. Concentration of RNA before microarray preparation.
7. Amplification, labelling and hybridisation of RNA
8. Scanning, data processing (normalisation, filtration)
9. Bioinformatic analyses to generate the list of differentially expressed genes between stable, BOS and RAS groups.
Proteomics analysis of blood and BAL (UJF)
10. Optimization of experimental conditions of plasma and BAL proteomic analysis. Validation of the analytical procedures
11. Sample preparation of blood samples with Proteominer® and BAL samples with ultrafiltration or dialysis.
12. SELDI-TOF and iTRAQ LC-MS/MS analysis performed on the sets of samples
13. Bioinformatics analysis done and establishement of a final list of dysregulated proteins between BOS and RAS phenotypes compared to stable phenotypes at 2 time points.
Characterization of LTR microbiome (CHUV)
14. We found that at 2 months, transplanted patients exhibited a markedly reduced bacterial load in comparison to non-transplanted, non-immunosuppressed and non-antibiotics-treated control subjects. At later time-points, the bacterial load tended to increase, although at a highly variable rate depending for a large part upon the infection status as defined by microbiological cultures and PCR assays (Institute of Microbiology, CHUV; See Annex - Figure 2.6).
15. This overall recolonization was accompanied by significant changes in phylum distribution. Samples up to 4 months post-transplantation displayed a predominance of Firmicutes (mean 38.7%) over Bacteroidetes (25.5%) and Proteobacteria (24.0% ; See Annex - Figure 2.7A). In contrast at 6 months and later, Bacteroidetes predominated (46.6%) over Firmicutes (26.9%) and Proteobacteria (17.0% ; See Annex - Figure 2.7A). Accordingly, the ratio of Bacteroidetes to Firmicutes showed a 4-fold median increase in samples obtained at 12-24 months, in comparison to samples from 0.5-2 months (See Annex - Figure 2.7B).
Hence, the data shows that the transplanted lung undergoes gradual recolonization with an overall shift from a Firmicutes- to Bacteroidetes-enriched microbial community.
Immunobiological assays (CHUN, CHUV)
16. The activation profile of LTR alveolar macrophages was analyzed using RT-qPCR. Principal component analysis (PCA) carried out on the resulting data set provided a distribution map used to delineate three sample groups that were further characterized (See Annex - Figure 2.8A). Based on a 2.9 and 11.6-fold median increase in TNF and COX-2 expression, respectively, as well as higher neutrophil counts (42% vs. 4%), we found a maximal association of inflammation with one specific sample group (See Annex - Figure 2.8B red dots). In contrast, tissue remodeling was assigned to a distinct sample group that displayed a 3.2-fold increase in PDGFD expression and a 2.6-fold higher TIMP1 to MMP12 expression ratio (See Annex - Figure 2.8C green dots). Hence, this characterization allowed us to discriminate between samples with polarized AM either with inflammation (12.9% of total sample set) or remodeling (46.8%) predominance, in accordance with the model of M1 and M2 activation profile, and to distinguish them from a sample group with non-polarized AM (40.3%; M0).
17. Time-dependent analysis of AM polarization in LTR indicated a slight predominance of M0 samples (54.5%) up to 2 months post-transplantation (See Annex - Figure 2.9). From 3 months and onwards, AM polarization prevailed, with increasing percentages of M2 samples over time, at the expense of M1 samples (57% and 7%, respectively, at 12 months). Taken together, these results point to the emergence of a subset of « long-term » LTR with alveolar conditions that promote macrophage polarization with high grade remodeling and low grade inflammation, that may lead to fibrosis and ultimately graft dysfunction. A longitudinal follow-up of these patients will define the prognostic value of this dynamic process and possible consequences on graft survival.
18. Finally, we found a potentially important link between BAL macrophage polarization and the phylum distribution in lung microbiota, as presented in section 3.4. Indeed, PCA of microbiota composition in samples classified as per AM gene expression profiling indicated only a partial overlap between the M0, M1 and M2 sample groups (See Annex - Figure 2.10A). As suggested by the correlation circle associated with PCA (See Annex - Figure 2.10B) a major characteristic of the microbiota composition in M2 samples with pro-remodeling AM profiling was the predominance of Bacteroidetes (median 54.0%; See Annex - Figure 2.10C). In contrast, this bacterial phylum was drastically under-represented in M1 samples with pro-inflammatory AM profiling (5.8%) and highly variable in M0 samples with non-polarized macrophages (27.4%). This suggests a potential association between the relative abundance of Bacteroidetes in lung microbiota and the underlying alveolar remodeling/inflammatory status.
Taken together, our results have allowed us to delineate 3 settings in the lower airways of lung transplant recipients: (i) a pro-fibrogenic AM profile combined with a Bacteroidetes-enriched microbiota, (ii) a pro-inflammatory AM profile combined with low Bacteroidetes, and (iii) a non-polarized AM gene expression profile potentially associated with a more balanced microbiota. Ongoing in-depth characterization of the lung microbiota based upon 16S rDNA sequencing analysis suggests that associations between a pro-remodeling AM profiling and particular species of Bacteroidetes can be identified. A patent application currently filed within this context represents an important deliverable linked to the present task. The assessment of a definitive link between these observations and long-term graft survival will only be made possible by the continuation of the intraindividual follow-up of this set of patients. However, we propose that the bronchoalveolar microbiota and the AM gene expression profile are interrelated in the transplanted lung and that a combined characterization of these two elements will help define the pathological conditions leading to CLAD occurrence.
WP 3: ESTABLISHMENT OF THE SYSCLAD KNOWLEDGE MANAGEMENT SYSTEM (BIOMAX) PARTNERS INVOLVED: CHUV, UJF, CHUN, ND, GATC, UZH
1. Established first CLAD specific information network integrating existing structured resources with literature mining results
2. Integrated clinical and multiple (6) type omics data from clinical and experimental centres and mapped it to the information network
3. Generated a CLAD research and therapy specific graphical user interface supporting clinical workflows and decision processes
4. Basic descriptive statistical analysis is provided
5. A CLAD specific, systematic literature analysis tool has been established and results are available
6. Alert tool is working
7. Harmonisation of clinicome is supported by the Information system
WP 4: SYSCLAD NUMERICAL DESIGN (ND) PARTNERS INVOLVED: CHUV, CHUN, UJF, BIOMAX
1. A clinical model has been successfully implemented that permits us to simulate the spirometry tests that are used in real clinical practice.
2. The simulation of the spirometry curves for a set of hypothetical healthy individuals was deemed to be compatible with reality by the SysCLAD consortium’s experts in the field.
3. The integration between the clinical and the biological computational models has permitted us to simulate the effects of any combination of perturbations (i.e. exposure to pollution, infectious episodes, etc.).
4. When running such simulations, it is possible now to analyse the evolution of a great amount of biological entities, in terms of their up-regulation or down-regulation.
WP 5: DISSEMINATION, ETHICS AND SOCIETAL ASPECTS (CHUV AND UJF)
PARTNERS INVOLVED: ALL PARTNERS
1. Efficient internal communication structure via sub-working groups,
2. Active targeted dissemination strategy, with over 70 presentations at scientific and non scientific events
3. Active follow up of ethical and gender issues of the project
WP 6: PROJECT AND INTELLECTUAL PROPERTY MANAGEMENT (CHUV/FINOVATIS) PARTNERS INVOLVED: ALL PARTNERS
1. All contractual agreements have been implemented at the start of the project. The pre-financing, received by CHUV was distributed to all partners who received 75% of their respective EC contribution. The management manual was circulated to all partners.
2. Monthly TCs implemented to follow up on technical progress and minutes circulated
3. An amendment request was submitted in February 2013 to proceed to the integration of University Zurich Hospital (UZH) and the group of Dr C. Benden in SysCLAD in order to significantly increase the number of patients recruited. The procedure was successful and UZH was part of SysCLAD with costs eligible from 1st of January 2013.
4. An amendment request was submitted in January 2014 to proceed to the integration of European Institute of Systems Biology and Medicine (EISBM, PI: C. Auffray) via the legal entity HLA MEDICINE. The procedure was successful and ESIBM was part of SysCLAD with costs eligible from 1st of January 2014.
Potential Impact:
The International Society for Heart and Lung Transplantation registry recorded that 3631 adult heart-lung transplants and 39 835 adult lung transplants have been performed in centres throughout the world up to 2012, establishing lung transplantation (LT) as the standard of care for selected patients with end-stage respiratory diseases. The long-term success of LT is limited by chronic lung allograft dysfunction (CLAD), of which the two most common phenotypes are an abnormal remodelling of the small airways resulting in progressive airflow obstruction called bronchiolitis obliterans syndrome (BOS) and, less frequently, a restrictive ventilatory process referred to as restrictive allograft syndrome (RAS). Almost 50% of LT recipients will develop CLAD within 5 years post-LT, rising to 75% after 10 years and represents the leading cause of deaths after year-1 post-LT.
In WP 1, the “clinicome” data are potentially one of the important contribution to the literature. Indeed clinical and functional data from the 422 patients which were reviewed after harmonization by successive adjudication committees , did allow to stratify patients into the sub phenotypes of CLAD as Stable, BOS, RAS or mixed, and others . The publication of the characteristic of each group will allow to improve the definitions of these entities in the literature, with the risk factors involved for them to develop and their ultimate prognosis. Prevention of some CLAD should be possible as soon as these risks will be validated. In addition, we will be better able to define their respective prognostic and avoid aggravating factors. The main data from the donors have been revised showing for instance that the donor age is not a real issue as long as no underlying lung disease is present. However we confirm that the age of the receiver is important as between 60 and 65 years old the overall survival is much poorer. These findings will increase the number of donor lung suitable for transplantation and will allow us to avoid to transplant too old patients with certain co-morbidities.
Immunophenotyping of PBMC has shown differential evolution of T cell phenotypes: an increase of interferon gamma producing T cells in patients developing CLAD, a decrease in central memory T cells and an increase in effector memory (EM) T cells in each pathological group with time. The EMRA cell population (the most differentiated type of memory cells) increased with time in BOS or RAS group, by contrast naive CD8+ cell population decreased .This demonstrates the feasibility and relevance of immunophenotyping for the monitoring of lung transplant recipients. These results will need to be validated by longer follow up and integrated into the prediction computationnel models of CLAD.
The data on the proteomics in the serum has brought new technology to the field of lung disease. The goal of this proteomic study was to identify at 6 and 12 months after LT, protein biomarkers that will predict CLAD occurrence 3 years later. As very few studies reported BAL analysis by iTRAQ-LC-MS/MS, alternative methods of sample preparation such as albumin depletion or other ultracentrifugation devices were extensively evaluated before settling the optimal experimental BAL conditions as previously described. It should be noted that this optimization study allowed us also to complete the description of BAL proteome from stable LTR to a level not even described in normal subjects (publication in progress). Regarding the CLAD biomarker investigation, 188 and 206 proteins were quantified in BAL at month 6 and 12, while 203 and 221 were quantified in plasma respectively. Among 244 proteins deregulated in BOS or RAS conditions, 6 were of particular interest for CLAD prediction. Analysis of plasma proteome by SELDI-TOF required a faster sample preparation than iTRAQ-LC-MS/MS analysis. Plasma samples needed to be enriched in low concentration proteins by the method of Proteominer® prior SELDI-TOF MS analysis. One hundred twenty seven to 300 protein peaks were identified depending on the sample set, and 6 to 65 proteins were found differentially expressed between groups (p-value<0.05). The best result was obtained by the comparison of BOS and ST plasmas at month 12, characterizing 52 protein peaks with an intensity above 2 and a p-value under 0.05. The non-hierarchical clustering of these biomarkers in a heatmap allowed the identification of the BOS patients with a specificity of 55.1% and a sensitivity of 88.2%. These data are promising to confirm or find new pathological pathways involved in CLAD . These are important steps forward and milestones.
Extraction of mRNA and microRNA was realized on whole blood collected after 6 and 12 months post transplantation (M6 and M12). RNA extraction provided good quality RNA since around 80% of RNA extracted fulfilled the requirements for microarrays. The differential expression of several genes of interest have been shown. Integration of these is underway to confirm the involvement of biologic cascades in different groups of CLAD.
All SysCLAD samples for exomes sequencing and InDels findings, fulfilled GATC's quality criteria. The identification of SNPs and InDels in most exons was extremely robust. For each sample 70,000 – 90,000 SNPs and 10,000 - 12,000 InDels exemplary data were identified. Of the InDels, 50% were insertions, the other half deletions. This raises hope that the SysCLAD consortium will identify SNPs and InDels involved in CLAD which were not described before. The best software solution to identify SNPs and InDels in a comparative analysis table was “Ingenuity Pathway Analysis” from Qiagen. For all SNPs already described in publications the software lists known associations with diseases or biological pathways. By comparing the different groups, highly significant SNPs or InDels were identified. These potential markers may allow discriminating the groups of interest. At the moment additional analyses are ongoing for these markers to identify variants which might be suitable for diagnostic purposes. The respective data tables will be made available to the consortium via the BioMax SysCLAD data base. This work is unique and correlations with the other omics is of great potential interest for building early diagnostic tools of CLAD by different SMEs . Genetic pharmacology might be envisaged for Interventions to prevent CLAD in personalized approaches.
Environmental pollution and potential impact on CLAD. Partnership agreement between Ineris (National Institute for Industrial Environment and Risks) and UJF Grenoble has been concluded in May 2014, and gave access to the up to date national-wide exposure model with a high spatial (1 x1 km) and temporal resolution. All data (mean, minimum and maximum daily exposures to PM2.5 PM10, NO2 and Ozone between October 2009 and December 2013) has been delivered by Ineris to UJF by end of june 2014 for patients. Metropolitans residents at the time of transplantation who were transplanted before July 2013, and who survived more than 6 months post-transplantant where shown to have lower levels of lung function related to the level of pollutants, but then similar decline in FEV1. Thus pollution seems to impact early recovery following lung transplant but less on the later decline. Increasing the collective and length of observation might show however that lung transplant might be more prone to CLAD and ultimately deaths.
Lung Microbote and lung inflammation post transplantation. SysCLAD did discover a time-dependent variations in the composition at phylum level of lung microbiota. The Comparative analysis of BAL samples obtained at "early" (0.5 to 4 months) or "late" (6 to 24 months) time-points post-transplantation shows that the transplanted lung undergoes gradual recolonization with an overall shift from a Firmicutes- to Bacteroidetes-enriched microbial community. Simultaneously time-dependent analysis of AM polarization in LTR indicated a slight predominance of M0 samples (54.5%) up to 2 months post-transplantation. From 3 months and onwards, AM polarization prevailed, with increasing percentages of M2 samples over time, at the expense of M1 samples (57% and 7%, respectively, at 12 months). These studies did discover thus contrasting changes in the relative abundance of M1 (inflammatory macrophages) vs. M2 (remodeling macrophages) samples over time post-transplantation. These results point to the emergence of a subset of « long-term » LTR with alveolar conditions that promote macrophage polarization with high grade remodeling and low grade inflammation, that may lead to fibrosis and ultimately graft dysfunction. In addition a potential association between the relative abundance of Bacteroidetes in lung microbiota and the underlying alveolar remodeling/inflammatory status is found. As gut and lung microbiota can be influenced by the type of food taken up by patients, if our findings are confirmed, diet recommendation could be made or antibiotic choosen accordingly .
Ongoing in-depth characterization of the lung microbiota based upon 16S rDNA sequencing analysis suggests that associations between a pro-remodeling AM profiling and particular species of Bacteroidetes can be identified. A patent application is currently filed for this novel finding which would lead thus to specific diagnostic and counseling approaches.
THE SYSCLAD KNOWLEDGE MANAGEMENT SYSTEM (BIOMAX)
Biomax did generate the first specific information network around CLAD, to bring together current general knowledge about human molecular process from existing resources. This information network integrate prior knowledge with disease-specific data, and this will be mined for connections of entities currently not known to participate in CLAD, by creating a transverse scientific network.
The Biomax platform did create the first CLAD research and potentially therapy-specific, user interface supporting clinical workflows and decision processes. This integration between currently disjointed literature based biomedical knowledge, molecular data and mathematical model is about to enable clinical researchers to apply system medicine workflows. Overall the first objective to provide a secure, federated software environment to integrate existing public and project-specific data sources and establish the data flow between project partners has been met.
The experimental data sets for all SysCLAD experimental methods (proteomics, UJF, m/miRNA transcriptomics CHUN, genotyping GATC, microbiome CHUV and immunobiological assays CHUN) have been integrated after step by step harmonization of data. Thus, the second objective to semantically map the different SysCLAD clinical and experimental data types, has been done being very helpful, and hallowing several communications in international meeting such as ISHLT in San Diego en 2014, in Paris during the 13th international meeting on lung transplantation and in the ISHLT of Nice in 2015, among other public manifestations. These will be fallowed by milestone publications
Lastly for the Knowledge model entities and references from Novadiscovery, upon which the knowledge management model was built, are included in the SysCLAD Information System of Biomax and connected to resources in the system. This has been made available with the data, but time has been short to be fully integrated and used at this stage, however is of great potential.
SYSCLAD NUMERICAL DESIGN BY NOVADISCOVERY AND EISBM
The main scientific innovation from this WP was expected to be an overall representation of all mechanisms involved in CLAD disease progression. In particular, new knowledge is expected to emerge in the form of new prognosis biomarkers and new treatment efficacy biomarkers.
Additional expected innovations are the structured accumulation & evaluation of knowledge available on the CLAD disease in the scientific literature through the development of a novel knowledge base and a graphical & comprehensive representation of the disease’s sub-mechanisms. The main expected technological innovations are : a CLAD risk prediction algorithm, which should take into account patient-specific descriptors. In particular, the development of a bedside personalised medicine web application.
The evaluation of the knowledge available on the CLAD disease is in revision for a publication in Transplantation journal and is therefore a step forward in the knowledge evolving CLAD and his two extreme phenotypes BOS and RAS.
The mathematical model has been built in great part despite the complexities of all parameters identified. One focus as clinical model, has permit us to simulate the spirometry tests. The integration between the clinical and the biological computational models has permitted to simulate the effects of any combination of perturbations (i.e. exposure to pollution, infectious episodes, etc.).
The models has been tested on data found in the litterature due to the late delivery of all omics data of SysCLAD due to the need of carefull harmonization of clinical and procedures used.
The late incoming of data and the complexities of all parameters involved has been difficult to integrate at this stage in the mathematical models. EISBM is providing SysCLAD with significant additional expertise in the integration of multi-level omics. In addition to data filtering and statistical analysis to qualify the large amount of data generated by the proteomics and additional transcriptomics experiments carried out by GATC (SNPs, InDels) that could feed the predictive model for both the calibration and validation phases in the future providing a follow up phase is found. Thus EISBM has been asked to use complementary system medicine technologies to integrate all omics and see what are the main biological pathways standing out in their objective analysis of available data in SysCLAD . Some specific molecules, for instance, seem to link proteomics and gene expressions found in CLAD, the same is through for some metalloproteinases. These analysis are still ongoing and major results are expected to be delivered in April for the ISHLT meeting in Nice and will surely be reported during the summer 2015.
DISSEMINATION, ETHICS AND SOCIETAL ASPECTS
The SysCLAD project was disseminated in various seminars, newsletters, technical papers etc . Several external communications channels were implemented. The public SysCLAD website (www.sysclad.eu) is online and freely accessible since M3. The website was updated on a monthly basis or when required (e.g. significant non-confidential milestone achieved). The partners have all been very active on the international scene to disseminate the project ambitions and ongoing results with presentations in more than 70 conferences/workshops/invited lectures and a publication in the European Parliament magazine.
All ethical issues have been solved before doing data analysis and reporting in major seminars. Only some timelag did occur for the role of environments on lung functions recovery and decay along the years.
The exploitation plan was subcontracted to the company TKM who has vast experience and a strong track record in the definition of business models and routes to market for technologies. TKM company has delivered and is contacting industrials stakeholders or organisations able to increase the visibility of SysCLAD and create new partnerships for the use of new biomarkers found or potential pathways to be targeted in view of novel therapies.
PROJECT AND INTELLECTUAL PROPERTY MANAGEMENT (CHUV/FINOVATIS)
Finovatis has been very instrumental for the implementation of a strong management structure to support the operational, technical and financial administration of the project. Identification of the best commercialisation strategy for the SysCLAD model in order to maximise technology transfer to clinicians and patients. The potential for a follow up of the SysCLAD cohort is being intensely scrutinized with the main partners of SysCLAD . Indeed many deliverables are promising to lead to new bridges between basic sciences discoveries and patient outcome management if all SysCLAD objectives could be implemented in the setting of system medicine.
List of Websites:
www.sysclad.eu
Recognized as the next major step forward in medical research, Systems Medicine (SM) offers a groundbreaking approach to understand and unravel complex diseases mechanisms and potentially overcome many of the current limitations of drug discovery. SM is also expected to lay the foundation of predictive, personalized, preventive and participatory (P4-Medicine), changing the way we practice medicine from a reactive to a proactive mode and focusing more on wellness rather than on disease. The FP7-funded research project SysCLAD, coordinated by Prof. L. Nicod, is one of the flagship initiatives of Systems Medicine in Europe, bringing together leading clinical centres in lung transplantation (LT), academic groups and industrials to i) identify and validate a Clinical and Molecular Signature Predictive of Chronic Allograft Lung Dysfunction (CLAD) and ii) develop a personalised medicine tool, based on mathematical representations of disease mechanisms, that will allow a prediction of CLAD resulting in Early and Specific Interventions LT patients at high risks to develop CLAD. The large knowledge base of integrated clinical/omics data produced for the validation of the SysCLAD model will be an invaluable resource for further SM-driven research in the context of other chronic respiratory diseases of major incidence such as Severe asthma, Chronic Obstructive Pulmonary Disease or Interstitial Pulmonary Fibrosis
The objectives of WP1 were: to obtain authorizations and ethical clearance from regulatory authorities from France, Belgium and Switzerland to merge COLT and STCS cohorts in order to recruit at least 840 candidates for lung transplantation and follow-up at least 400 LTR for 3 years; to define and disseminate Standard operating procedures (SOPs); to store and deliver samples to other WP; to guarantee the quality of data generated in the project. All authorizations and ethical clearance were obtained, with some significant delay for CNIL clearance on the environmental exposure study. All SOPs were disseminated to all participating centres. The objective of 840 candidates for lung transplantation has been exceeded and 422 LTR have now been follow-up during 3 years. Clinical and functional data from these 422 patients were reviewed and harmonised by successive adjudication committees in order to stratify patients into the sub phenotypes of CLAD as Stable, BOS, RAS, mixed or other. High quality biological samples and data from these patients at M6 and M12 post-transplantation were delivered to partners for omics, microbiome and immunological analysis to start the planned analysis. The only deviation from the original work plan was linked to the environmental exposure study which was started later than planned due to delayed authorisations. However, all work was carried out by UJF to correct this delay and UJF provided accurate estimates of the chronic exposure to ambient air pollutant exposure at the individual level and assessed the association between chronic air pollutant exposure and FEV1 level and decline.
In WP2, Samples from the SysCLAD biobank (blood, BAL) were analysed by a wide range of analytical methods covering transcriptomics (mRNA, miRNA), full exome sequencing, proteomics (SELDI-TOF, iTRAQ LC-MS/MS), immunological bioassays and microbiome characterisation according to the various CLAD sub-phenotypes identified in WP1. All data were fed into the global database generated by BIOMAX before complete integration alongside the clinical parameters into the numerical model. This WP led to the identification of signatures of each CLAD sub-phenotypes according to differences in genes and proteins expression in lung transplant recipients, and allowed the identification of specific patterns of pulmonary microbiota composition varying in concert with the underlying alveolar inflammatory background. In WP3, a stable SysCLAD Information System was created, where clinical, experimental and data from external databases are integrated and linked. A dictionary-based literature analysis tool with graphical integration of the results adapted to the needs of the clinician community was developed and disseminated.
In WP4, the computational model developed by Novadiscovery in the context of the SysCLAD project comprises three fundamental bricks: i) a disease model, which describes, computationally, the pathogenesis of CLAD; ii) a clinical model, which translates the outcome of the disease model into a real clinical variable, such as those used to assess the pulmonary function of a patient; iii) a virtual population, which describes the inter-individual variability that characterizes the natural history of CLAD.
In summary, the main objective for this work package was not fully achieved. However:
• The mathematical model is not far from being completed.
• Instead, we developed a logic-based model (it was not initially planned) that enabled to achieve some of the objectives.
• We developed a software application to simulate any combination of perturbations (pollutants, infections, GERD, etc…)
• We developed a Virtual Population with unrestricted values for patients’ descriptors.
The framework established in WP5 ensured efficient exchange of scientific information both within the consortium and to the wider scientific and industrial communities through a vast number of dissemination activities to all stakeholders via the numerous communication channels available to the partners and establishing new ones. In addition, it allowed an efficient follow-up and management of ethical, regulatory and societal issues that arised during the project.
All procedures for internal communication, technical and financial reporting/IP monitoring at WP level, partner level and consortium wide were efficiently implemented. SysCLAD governance was extremely efficient to rapidly take and act upon consortium decisions. Procedures were specifically developed to be adapted to the nature of the SysCLAD project (i.e. SME-centric) allowing for a thorough yet agile management and follow up of all activities of the project. Monthly TCs, led by FINOVATIS were implemented to coordinate technical, financial and administrative activities and identify any potential issues and define corrective actions in a rapid manner. No significant deviations from the original work plan were notable from a management standpoint.
Project Context and Objectives:
With more than 32,000 procedures performed around the world in the least 30 years, lung transplantation (LT) has become the standard of care for selected patients with advanced lung diseases. Unfortunately, LT procedures are often at risk of chronic lung allograft dysfunction (CLAD), of which the two most common pathologic presentations are i) a remodelling of small airways resulting in an obstructive ventilatory pattern “Bronchiolitis Obliterans Syndrome” (BOS) and ii) a restrictive ventilatory pattern “Restrictive Allograft Syndrome” (RAS). Almost 50% of LT recipients will develop CLAD within 5 years post LT, rising to 75% after 10 years. The median survival figures from the time of diagnosis remain very poor for both early-onset CLAD (1.5 years) and late-onset CLAD (2.5 years). In clinical routine, the decline of the forced expiratory volume in one second (FEV1) is used as a first indication of CLAD onset once differential diagnoses (e.g. acute rejection, active infection and/or proximal bronchial complications) have been ruled out. Treating CLAD remains a challenge for several reasons: the main pathways, molecular events and triggers leading to Epithelial-Mesenchymal Transition (EMT) within the small airways or transformation of mesenchymal stromal cells, are not well understood and there has been, to date, no real attempts, due to the lack of tools, to integrate risks factors for a specific lung recipient to anticipate lung function decline before irreversible lung function loss. In the absence of robust guidelines and compounds active on EMT, CLAD treatment is often initiated late, chaotic and involves an increase in immunosuppressant which often leads to associated opportunistic infections, kidney failure, re-transplantation and, ultimately, death. An early prediction of CLAD based on principles of systems biology and personalized medicine would represent a major breakthrough and open a new area marked by early interventions based on risk stratification and personalized pathways modulation. Early recognition and specific intervention in patients at risk of CLAD would undoubtedly improve outcomes.
The major aim of this project is to build the first SysCLAD computational model to be used by clinicians and researchers. The model will identify and validate a “signature” of CLAD integrating multilevel data in order to i) identify and mitigate the risk of chronic allograft dysfunction and ii) allow for early intervention before irreversible damage. The SysCLAD computational model created in this project should allow prediction within the first year post LT of which recipients are at risk to develop BOS or RAS by 3 years post LT. This prediction, based on a mathematical and computer model developed through a systems biology approach, will be derived from the complete integration of large experimental data (clinicome, environmental data, omics, microbiome, and immunological assays) collected from both donors and recipients (See Annex - Supporting Fig. 1.1). The project will establish a prospective cohort of LT candidates (European Cohort Of Lung Transplantation (ECOLT)), which will expand the ongoing Cohort Of Lung Transplantation (COLT) (http://www.clinicaltrial.gov/ClinicalTrials.gov Identifier: NCT00980967) recruited since June 2009.
The SysCLAD model is constructed using data from a first cohort of 200 LT recipients and refined/validated using the new set of 200 LT data generated by ECOLT. In order to develop the SysCLAD model, the following five objectives are pursued in an integrated workflow:
1. Recruitment of LT candidates and LT recipients (ECOLT) was initiated in November 2012: 500 LTR were already included between June 2009 and January 2012. In addition to this, the inclusion of 200 new candidates (surviving more than 3 months post LT) resulted in availability of more than 700 LTR by January 2013 at project start. Of these, 400 LTR will reach 3 years follow up by July 2014 (M18) and be used to build, calibrate and validate the SysCLAD prediction model. In parallel, standard operating procedures related to sample collection (timing, type and mode of data collection) that have already been established will be extended to others members of the consortium for global data harmonisation.
2. Proteomics, transcriptomics and immunological assays of minimally invasive sampled fluids (blood and bronchoalveolar lavage - BAL) will be carried out and used to achieve objective 3.
3. A comprehensive clinical research information system (CRIS) centralising all the clinical and experimental data (environment, phenotype, microbiome, biology, omics) collected from donors, recipients in a longitudinal fashion will be built.
4. CLAD signature profiling. The data generated from objective 3, together with additional data from the literature will be integrated and modelled in order to build the first signature of CLAD.
5. Validation of SysCLAD. The SysCLAD model will be refined and validated by iteration in consecutive sets (see figure below).
To yield substantial benefits to drug discovery and development efforts, disease modelling needs to embrace the full spectrum of mechanisms and their interactions, from genes to populations. Our approach, which purports to design multi-scale mathematical disease models plugged in virtual populations, represents a powerful link to bridge systems biology and clinical applications together for the ultimate benefit of clinicians and patients.
The overarching objective of SysCLAD is to deliver personalised medicine and drug comparative-effectiveness prediction capabilities over a specifically characterised target population. The systems biology-driven modelling approach includes all the biological components that are thought to play a role in the course of a disease process, up to the clinical events, by plugging the disease model into the virtual population. The virtual population represents the population of interest. Each virtual patient of the population is characterised by a vector of descriptors that translate biological and environmental parameters involved in the course of the disease. Applied to CLAD modelling, the virtual population will be created from environmental, clinical and molecular vectors.
The specific objectives of each scientific work package were as follows:
WP 1: SOPs, ECOLT COHORT DESIGN, BIOBANKING (CHUN)
• To edict and set up Standard Operating Procedures (SOPs) adapted to Lung Transplantation in the 13 participating Centres of Lung Transplantation.
• To recruit 840 candidates for lung transplantation from whom 700 LTR
• To follow-up 400 Lung Transplant recipients for 3 years
• To constitute a database of clinical and functional data from 400 LTR
• To constitute a certificated biobank from 400 Lung Transplant recipients
• To deliver data and samples to investigators from WP 2 -5 as 2 successive cohorts of 200 patients with a 3 years follow-up
• To characterise the environmental exposure of LTR
WP 2: GENERATION OF BIOMARKERS TO BUILD SYSCLAD (UJF)
This WP was specifically dedicated to the generation of experimental data for the construction and the validation of the SysCLAD model. It is constituted of several parts including blood transcriptomics, bronchoalveolar lavage (BAL) and blood proteomics, recipient polymorphism analysis on blood, analysis of LTR lung microbiomes, and immunobiological assays on blood and BAL. The results of this WP allow the definition of transcriptomic, proteomic and immunobiological profiles of CLAD that can be used to build the SysCLAD model. In addition, these results will lead to the identification of new biomarkers for CLAD prediction and allow a better understanding of CLAD disease mechanisms.
WP 3: ESTABLISHMENT OF THE SYSCLAD KNOWLEDGE MANAGEMENT SYSTEM (BIOMAX)
Work package 3 will establish the content and technical infrastructure necessary to connect the clinical and experimental data generated within WP1 and 2 with the SysCLAD model developed in WP4. Specific objectives are:
• To provide a secure, federated software environment to integrate existing public and project-specific data sources and establish the data flow between project partners
• To semantically map the different SysCLAD clinical and experimental data types, information from public repositories and knowledge extracted from the literature
WP 3 will generate the first specific information network around CLAD, enabling WP4 to mine the molecular mechanisms involved in the process of developing BOS/EMT. It will bring together current general knowledge about human molecular process from existing resources such as KEGG with CLAD-specific, but unstructured knowledge available in the literature and available clinical and experimental data. This information network integrating prior knowledge with disease-specific data will be mined for connections and entities currently not known to participate in CLAD or involved in new roles. Technologically, WP3 will create the first CLAD research and therapy-specific graphical user interface supporting clinical workflows and decision processes. Technical integration between currently disjointed CRIS, literature based biomedical knowledge, molecular data and mathematical model will enable clinical researchers to apply the system medicine workflow. An existing literature mining process will be adapted to CLAD-specific information e.g. through development of a dictionary of CLAD-related terms for literature mining. Finally an existing knowledge management technology will be adapted to new experimental technologies and corresponding data types and formats.
WP 4: SYSCLAD NUMERICAL DESIGN (ND)
The primary WP4 objective was to develop a fully operational numerical model of CLAD to enable patient-level prediction of CLAD occurrence. At the beginning of the project, it was thought that the global model would consist in 6 sub-models:, i) Non-allo-dependent inflammation (prolonged ischemia, ischemia-reperfusion lesion, lymphatic drain disruption, denervation, gastro oesophagus reflux, etc.); ii) Allo and autoimmune mechanisms; iii) Infections (viral: CMV, rhinovirus, etc.; bacterial); iv) Innervation (without the structural and organisational aspects) with mediators, stimuli and targets linking to the nervous system; v) Model of the bronchiolitis epithelium; vi) Functional anatomy of bronchioles (from a phenomenological standpoint): stenosis growth and its impact on airway flow, which would yield the clinical criteria to stratify patients (a quantitative criterion translated into a binary “>0 or <0” variable).
The main scientific innovation from this WP was expected to be an overall representation of all mechanisms involved in CLAD disease progression. In particular, new knowledge was expected to emerge in the form of new prognosis biomarkers and new treatment efficacy biomarkers. Additional expected innovations were the structured accumulation & evaluation of knowledge available on the CLAD disease in the scientific literature through the development of a novel knowledge base and a graphical & comprehensive representation of the disease’s sub-mechanisms. The main expected technological innovations were: a CLAD risk prediction algorithm, which should take into account patient-specific descriptors. In particular, the development of a bedside personalised medicine web application was planned, for clinicians to (i) predict the likelihood of CLAD occurrence and (ii) predict optimal treatment responses. On the longer term, the CLAD numerical model associated with its virtual population should enable new potential therapeutic targets to be identified for new therapy development.
Project Results:
Significant results and achievements from each work package of SysCLAD
WP 1: SOPS, ECOLT COHORT DESIGN, BIOBANKING (CHUN) PARTNERS INVOLVED: CHUV, UJF, UZH
1. Implementation of Standard Operating Procedures (SOPs) adapted to Lung Transplantation in the 14 participating Centres of Lung Transplantation.
2. Recruitment of 1608 candidates for lung transplantation of which 1107 are LTR
3. Follow-up 422 LTR for 3 years
4. Development of a clinical and functional harmonised database for 422 LTR, with Biomax partner, freely accessible to all partners
5. Biobanking Certification obtained for samples storage
6. high quality biological samples and associated data collected from the well characterized 220 SysCLAD LTR patients have been delivered to investigators from WP 2 -5
7. Accurate estimates of chronic exposure to ambient air pollutant exposure at the individual level
8. First study suggesting an association between chronic exposure to ambient air pollutants and FEV1 level in LT patients
WP 2: GENERATION OF BIOMARKERS TO BUILD SYSCLAD (UJF) PARTNERS INVOLVED: CHUV, CHUN, GATC
Exome analysis (GATC)
1. Evaluation of all exon sequencing kits. Agilent’s SureSelect All Exon V5 have been used for the analysis of exome sequencing samples provided by SysClad.
2. Both Illumina’s Sequencing Systems HiSeq 2000 and MiSeq were suitable and were utilized for sequencing all exon samples.
3. Several possibilities to do comparative analysis of SNPs and InDels were evaluated to identify the most appropriate approach. GATC has decided to use a third party web-based analysis tool (see also “deviations”).
4. All samples were successfully sequenced and analysed.
Transcriptomic mRNA and miRNA Analysis (CHUN)
5. Extraction of RNA and RNA quality verification (>80% passed)
6. Concentration of RNA before microarray preparation.
7. Amplification, labelling and hybridisation of RNA
8. Scanning, data processing (normalisation, filtration)
9. Bioinformatic analyses to generate the list of differentially expressed genes between stable, BOS and RAS groups.
Proteomics analysis of blood and BAL (UJF)
10. Optimization of experimental conditions of plasma and BAL proteomic analysis. Validation of the analytical procedures
11. Sample preparation of blood samples with Proteominer® and BAL samples with ultrafiltration or dialysis.
12. SELDI-TOF and iTRAQ LC-MS/MS analysis performed on the sets of samples
13. Bioinformatics analysis done and establishement of a final list of dysregulated proteins between BOS and RAS phenotypes compared to stable phenotypes at 2 time points.
Characterization of LTR microbiome (CHUV)
14. We found that at 2 months, transplanted patients exhibited a markedly reduced bacterial load in comparison to non-transplanted, non-immunosuppressed and non-antibiotics-treated control subjects. At later time-points, the bacterial load tended to increase, although at a highly variable rate depending for a large part upon the infection status as defined by microbiological cultures and PCR assays (Institute of Microbiology, CHUV; See Annex - Figure 2.6).
15. This overall recolonization was accompanied by significant changes in phylum distribution. Samples up to 4 months post-transplantation displayed a predominance of Firmicutes (mean 38.7%) over Bacteroidetes (25.5%) and Proteobacteria (24.0% ; See Annex - Figure 2.7A). In contrast at 6 months and later, Bacteroidetes predominated (46.6%) over Firmicutes (26.9%) and Proteobacteria (17.0% ; See Annex - Figure 2.7A). Accordingly, the ratio of Bacteroidetes to Firmicutes showed a 4-fold median increase in samples obtained at 12-24 months, in comparison to samples from 0.5-2 months (See Annex - Figure 2.7B).
Hence, the data shows that the transplanted lung undergoes gradual recolonization with an overall shift from a Firmicutes- to Bacteroidetes-enriched microbial community.
Immunobiological assays (CHUN, CHUV)
16. The activation profile of LTR alveolar macrophages was analyzed using RT-qPCR. Principal component analysis (PCA) carried out on the resulting data set provided a distribution map used to delineate three sample groups that were further characterized (See Annex - Figure 2.8A). Based on a 2.9 and 11.6-fold median increase in TNF and COX-2 expression, respectively, as well as higher neutrophil counts (42% vs. 4%), we found a maximal association of inflammation with one specific sample group (See Annex - Figure 2.8B red dots). In contrast, tissue remodeling was assigned to a distinct sample group that displayed a 3.2-fold increase in PDGFD expression and a 2.6-fold higher TIMP1 to MMP12 expression ratio (See Annex - Figure 2.8C green dots). Hence, this characterization allowed us to discriminate between samples with polarized AM either with inflammation (12.9% of total sample set) or remodeling (46.8%) predominance, in accordance with the model of M1 and M2 activation profile, and to distinguish them from a sample group with non-polarized AM (40.3%; M0).
17. Time-dependent analysis of AM polarization in LTR indicated a slight predominance of M0 samples (54.5%) up to 2 months post-transplantation (See Annex - Figure 2.9). From 3 months and onwards, AM polarization prevailed, with increasing percentages of M2 samples over time, at the expense of M1 samples (57% and 7%, respectively, at 12 months). Taken together, these results point to the emergence of a subset of « long-term » LTR with alveolar conditions that promote macrophage polarization with high grade remodeling and low grade inflammation, that may lead to fibrosis and ultimately graft dysfunction. A longitudinal follow-up of these patients will define the prognostic value of this dynamic process and possible consequences on graft survival.
18. Finally, we found a potentially important link between BAL macrophage polarization and the phylum distribution in lung microbiota, as presented in section 3.4. Indeed, PCA of microbiota composition in samples classified as per AM gene expression profiling indicated only a partial overlap between the M0, M1 and M2 sample groups (See Annex - Figure 2.10A). As suggested by the correlation circle associated with PCA (See Annex - Figure 2.10B) a major characteristic of the microbiota composition in M2 samples with pro-remodeling AM profiling was the predominance of Bacteroidetes (median 54.0%; See Annex - Figure 2.10C). In contrast, this bacterial phylum was drastically under-represented in M1 samples with pro-inflammatory AM profiling (5.8%) and highly variable in M0 samples with non-polarized macrophages (27.4%). This suggests a potential association between the relative abundance of Bacteroidetes in lung microbiota and the underlying alveolar remodeling/inflammatory status.
Taken together, our results have allowed us to delineate 3 settings in the lower airways of lung transplant recipients: (i) a pro-fibrogenic AM profile combined with a Bacteroidetes-enriched microbiota, (ii) a pro-inflammatory AM profile combined with low Bacteroidetes, and (iii) a non-polarized AM gene expression profile potentially associated with a more balanced microbiota. Ongoing in-depth characterization of the lung microbiota based upon 16S rDNA sequencing analysis suggests that associations between a pro-remodeling AM profiling and particular species of Bacteroidetes can be identified. A patent application currently filed within this context represents an important deliverable linked to the present task. The assessment of a definitive link between these observations and long-term graft survival will only be made possible by the continuation of the intraindividual follow-up of this set of patients. However, we propose that the bronchoalveolar microbiota and the AM gene expression profile are interrelated in the transplanted lung and that a combined characterization of these two elements will help define the pathological conditions leading to CLAD occurrence.
WP 3: ESTABLISHMENT OF THE SYSCLAD KNOWLEDGE MANAGEMENT SYSTEM (BIOMAX) PARTNERS INVOLVED: CHUV, UJF, CHUN, ND, GATC, UZH
1. Established first CLAD specific information network integrating existing structured resources with literature mining results
2. Integrated clinical and multiple (6) type omics data from clinical and experimental centres and mapped it to the information network
3. Generated a CLAD research and therapy specific graphical user interface supporting clinical workflows and decision processes
4. Basic descriptive statistical analysis is provided
5. A CLAD specific, systematic literature analysis tool has been established and results are available
6. Alert tool is working
7. Harmonisation of clinicome is supported by the Information system
WP 4: SYSCLAD NUMERICAL DESIGN (ND) PARTNERS INVOLVED: CHUV, CHUN, UJF, BIOMAX
1. A clinical model has been successfully implemented that permits us to simulate the spirometry tests that are used in real clinical practice.
2. The simulation of the spirometry curves for a set of hypothetical healthy individuals was deemed to be compatible with reality by the SysCLAD consortium’s experts in the field.
3. The integration between the clinical and the biological computational models has permitted us to simulate the effects of any combination of perturbations (i.e. exposure to pollution, infectious episodes, etc.).
4. When running such simulations, it is possible now to analyse the evolution of a great amount of biological entities, in terms of their up-regulation or down-regulation.
WP 5: DISSEMINATION, ETHICS AND SOCIETAL ASPECTS (CHUV AND UJF)
PARTNERS INVOLVED: ALL PARTNERS
1. Efficient internal communication structure via sub-working groups,
2. Active targeted dissemination strategy, with over 70 presentations at scientific and non scientific events
3. Active follow up of ethical and gender issues of the project
WP 6: PROJECT AND INTELLECTUAL PROPERTY MANAGEMENT (CHUV/FINOVATIS) PARTNERS INVOLVED: ALL PARTNERS
1. All contractual agreements have been implemented at the start of the project. The pre-financing, received by CHUV was distributed to all partners who received 75% of their respective EC contribution. The management manual was circulated to all partners.
2. Monthly TCs implemented to follow up on technical progress and minutes circulated
3. An amendment request was submitted in February 2013 to proceed to the integration of University Zurich Hospital (UZH) and the group of Dr C. Benden in SysCLAD in order to significantly increase the number of patients recruited. The procedure was successful and UZH was part of SysCLAD with costs eligible from 1st of January 2013.
4. An amendment request was submitted in January 2014 to proceed to the integration of European Institute of Systems Biology and Medicine (EISBM, PI: C. Auffray) via the legal entity HLA MEDICINE. The procedure was successful and ESIBM was part of SysCLAD with costs eligible from 1st of January 2014.
Potential Impact:
The International Society for Heart and Lung Transplantation registry recorded that 3631 adult heart-lung transplants and 39 835 adult lung transplants have been performed in centres throughout the world up to 2012, establishing lung transplantation (LT) as the standard of care for selected patients with end-stage respiratory diseases. The long-term success of LT is limited by chronic lung allograft dysfunction (CLAD), of which the two most common phenotypes are an abnormal remodelling of the small airways resulting in progressive airflow obstruction called bronchiolitis obliterans syndrome (BOS) and, less frequently, a restrictive ventilatory process referred to as restrictive allograft syndrome (RAS). Almost 50% of LT recipients will develop CLAD within 5 years post-LT, rising to 75% after 10 years and represents the leading cause of deaths after year-1 post-LT.
In WP 1, the “clinicome” data are potentially one of the important contribution to the literature. Indeed clinical and functional data from the 422 patients which were reviewed after harmonization by successive adjudication committees , did allow to stratify patients into the sub phenotypes of CLAD as Stable, BOS, RAS or mixed, and others . The publication of the characteristic of each group will allow to improve the definitions of these entities in the literature, with the risk factors involved for them to develop and their ultimate prognosis. Prevention of some CLAD should be possible as soon as these risks will be validated. In addition, we will be better able to define their respective prognostic and avoid aggravating factors. The main data from the donors have been revised showing for instance that the donor age is not a real issue as long as no underlying lung disease is present. However we confirm that the age of the receiver is important as between 60 and 65 years old the overall survival is much poorer. These findings will increase the number of donor lung suitable for transplantation and will allow us to avoid to transplant too old patients with certain co-morbidities.
Immunophenotyping of PBMC has shown differential evolution of T cell phenotypes: an increase of interferon gamma producing T cells in patients developing CLAD, a decrease in central memory T cells and an increase in effector memory (EM) T cells in each pathological group with time. The EMRA cell population (the most differentiated type of memory cells) increased with time in BOS or RAS group, by contrast naive CD8+ cell population decreased .This demonstrates the feasibility and relevance of immunophenotyping for the monitoring of lung transplant recipients. These results will need to be validated by longer follow up and integrated into the prediction computationnel models of CLAD.
The data on the proteomics in the serum has brought new technology to the field of lung disease. The goal of this proteomic study was to identify at 6 and 12 months after LT, protein biomarkers that will predict CLAD occurrence 3 years later. As very few studies reported BAL analysis by iTRAQ-LC-MS/MS, alternative methods of sample preparation such as albumin depletion or other ultracentrifugation devices were extensively evaluated before settling the optimal experimental BAL conditions as previously described. It should be noted that this optimization study allowed us also to complete the description of BAL proteome from stable LTR to a level not even described in normal subjects (publication in progress). Regarding the CLAD biomarker investigation, 188 and 206 proteins were quantified in BAL at month 6 and 12, while 203 and 221 were quantified in plasma respectively. Among 244 proteins deregulated in BOS or RAS conditions, 6 were of particular interest for CLAD prediction. Analysis of plasma proteome by SELDI-TOF required a faster sample preparation than iTRAQ-LC-MS/MS analysis. Plasma samples needed to be enriched in low concentration proteins by the method of Proteominer® prior SELDI-TOF MS analysis. One hundred twenty seven to 300 protein peaks were identified depending on the sample set, and 6 to 65 proteins were found differentially expressed between groups (p-value<0.05). The best result was obtained by the comparison of BOS and ST plasmas at month 12, characterizing 52 protein peaks with an intensity above 2 and a p-value under 0.05. The non-hierarchical clustering of these biomarkers in a heatmap allowed the identification of the BOS patients with a specificity of 55.1% and a sensitivity of 88.2%. These data are promising to confirm or find new pathological pathways involved in CLAD . These are important steps forward and milestones.
Extraction of mRNA and microRNA was realized on whole blood collected after 6 and 12 months post transplantation (M6 and M12). RNA extraction provided good quality RNA since around 80% of RNA extracted fulfilled the requirements for microarrays. The differential expression of several genes of interest have been shown. Integration of these is underway to confirm the involvement of biologic cascades in different groups of CLAD.
All SysCLAD samples for exomes sequencing and InDels findings, fulfilled GATC's quality criteria. The identification of SNPs and InDels in most exons was extremely robust. For each sample 70,000 – 90,000 SNPs and 10,000 - 12,000 InDels exemplary data were identified. Of the InDels, 50% were insertions, the other half deletions. This raises hope that the SysCLAD consortium will identify SNPs and InDels involved in CLAD which were not described before. The best software solution to identify SNPs and InDels in a comparative analysis table was “Ingenuity Pathway Analysis” from Qiagen. For all SNPs already described in publications the software lists known associations with diseases or biological pathways. By comparing the different groups, highly significant SNPs or InDels were identified. These potential markers may allow discriminating the groups of interest. At the moment additional analyses are ongoing for these markers to identify variants which might be suitable for diagnostic purposes. The respective data tables will be made available to the consortium via the BioMax SysCLAD data base. This work is unique and correlations with the other omics is of great potential interest for building early diagnostic tools of CLAD by different SMEs . Genetic pharmacology might be envisaged for Interventions to prevent CLAD in personalized approaches.
Environmental pollution and potential impact on CLAD. Partnership agreement between Ineris (National Institute for Industrial Environment and Risks) and UJF Grenoble has been concluded in May 2014, and gave access to the up to date national-wide exposure model with a high spatial (1 x1 km) and temporal resolution. All data (mean, minimum and maximum daily exposures to PM2.5 PM10, NO2 and Ozone between October 2009 and December 2013) has been delivered by Ineris to UJF by end of june 2014 for patients. Metropolitans residents at the time of transplantation who were transplanted before July 2013, and who survived more than 6 months post-transplantant where shown to have lower levels of lung function related to the level of pollutants, but then similar decline in FEV1. Thus pollution seems to impact early recovery following lung transplant but less on the later decline. Increasing the collective and length of observation might show however that lung transplant might be more prone to CLAD and ultimately deaths.
Lung Microbote and lung inflammation post transplantation. SysCLAD did discover a time-dependent variations in the composition at phylum level of lung microbiota. The Comparative analysis of BAL samples obtained at "early" (0.5 to 4 months) or "late" (6 to 24 months) time-points post-transplantation shows that the transplanted lung undergoes gradual recolonization with an overall shift from a Firmicutes- to Bacteroidetes-enriched microbial community. Simultaneously time-dependent analysis of AM polarization in LTR indicated a slight predominance of M0 samples (54.5%) up to 2 months post-transplantation. From 3 months and onwards, AM polarization prevailed, with increasing percentages of M2 samples over time, at the expense of M1 samples (57% and 7%, respectively, at 12 months). These studies did discover thus contrasting changes in the relative abundance of M1 (inflammatory macrophages) vs. M2 (remodeling macrophages) samples over time post-transplantation. These results point to the emergence of a subset of « long-term » LTR with alveolar conditions that promote macrophage polarization with high grade remodeling and low grade inflammation, that may lead to fibrosis and ultimately graft dysfunction. In addition a potential association between the relative abundance of Bacteroidetes in lung microbiota and the underlying alveolar remodeling/inflammatory status is found. As gut and lung microbiota can be influenced by the type of food taken up by patients, if our findings are confirmed, diet recommendation could be made or antibiotic choosen accordingly .
Ongoing in-depth characterization of the lung microbiota based upon 16S rDNA sequencing analysis suggests that associations between a pro-remodeling AM profiling and particular species of Bacteroidetes can be identified. A patent application is currently filed for this novel finding which would lead thus to specific diagnostic and counseling approaches.
THE SYSCLAD KNOWLEDGE MANAGEMENT SYSTEM (BIOMAX)
Biomax did generate the first specific information network around CLAD, to bring together current general knowledge about human molecular process from existing resources. This information network integrate prior knowledge with disease-specific data, and this will be mined for connections of entities currently not known to participate in CLAD, by creating a transverse scientific network.
The Biomax platform did create the first CLAD research and potentially therapy-specific, user interface supporting clinical workflows and decision processes. This integration between currently disjointed literature based biomedical knowledge, molecular data and mathematical model is about to enable clinical researchers to apply system medicine workflows. Overall the first objective to provide a secure, federated software environment to integrate existing public and project-specific data sources and establish the data flow between project partners has been met.
The experimental data sets for all SysCLAD experimental methods (proteomics, UJF, m/miRNA transcriptomics CHUN, genotyping GATC, microbiome CHUV and immunobiological assays CHUN) have been integrated after step by step harmonization of data. Thus, the second objective to semantically map the different SysCLAD clinical and experimental data types, has been done being very helpful, and hallowing several communications in international meeting such as ISHLT in San Diego en 2014, in Paris during the 13th international meeting on lung transplantation and in the ISHLT of Nice in 2015, among other public manifestations. These will be fallowed by milestone publications
Lastly for the Knowledge model entities and references from Novadiscovery, upon which the knowledge management model was built, are included in the SysCLAD Information System of Biomax and connected to resources in the system. This has been made available with the data, but time has been short to be fully integrated and used at this stage, however is of great potential.
SYSCLAD NUMERICAL DESIGN BY NOVADISCOVERY AND EISBM
The main scientific innovation from this WP was expected to be an overall representation of all mechanisms involved in CLAD disease progression. In particular, new knowledge is expected to emerge in the form of new prognosis biomarkers and new treatment efficacy biomarkers.
Additional expected innovations are the structured accumulation & evaluation of knowledge available on the CLAD disease in the scientific literature through the development of a novel knowledge base and a graphical & comprehensive representation of the disease’s sub-mechanisms. The main expected technological innovations are : a CLAD risk prediction algorithm, which should take into account patient-specific descriptors. In particular, the development of a bedside personalised medicine web application.
The evaluation of the knowledge available on the CLAD disease is in revision for a publication in Transplantation journal and is therefore a step forward in the knowledge evolving CLAD and his two extreme phenotypes BOS and RAS.
The mathematical model has been built in great part despite the complexities of all parameters identified. One focus as clinical model, has permit us to simulate the spirometry tests. The integration between the clinical and the biological computational models has permitted to simulate the effects of any combination of perturbations (i.e. exposure to pollution, infectious episodes, etc.).
The models has been tested on data found in the litterature due to the late delivery of all omics data of SysCLAD due to the need of carefull harmonization of clinical and procedures used.
The late incoming of data and the complexities of all parameters involved has been difficult to integrate at this stage in the mathematical models. EISBM is providing SysCLAD with significant additional expertise in the integration of multi-level omics. In addition to data filtering and statistical analysis to qualify the large amount of data generated by the proteomics and additional transcriptomics experiments carried out by GATC (SNPs, InDels) that could feed the predictive model for both the calibration and validation phases in the future providing a follow up phase is found. Thus EISBM has been asked to use complementary system medicine technologies to integrate all omics and see what are the main biological pathways standing out in their objective analysis of available data in SysCLAD . Some specific molecules, for instance, seem to link proteomics and gene expressions found in CLAD, the same is through for some metalloproteinases. These analysis are still ongoing and major results are expected to be delivered in April for the ISHLT meeting in Nice and will surely be reported during the summer 2015.
DISSEMINATION, ETHICS AND SOCIETAL ASPECTS
The SysCLAD project was disseminated in various seminars, newsletters, technical papers etc . Several external communications channels were implemented. The public SysCLAD website (www.sysclad.eu) is online and freely accessible since M3. The website was updated on a monthly basis or when required (e.g. significant non-confidential milestone achieved). The partners have all been very active on the international scene to disseminate the project ambitions and ongoing results with presentations in more than 70 conferences/workshops/invited lectures and a publication in the European Parliament magazine.
All ethical issues have been solved before doing data analysis and reporting in major seminars. Only some timelag did occur for the role of environments on lung functions recovery and decay along the years.
The exploitation plan was subcontracted to the company TKM who has vast experience and a strong track record in the definition of business models and routes to market for technologies. TKM company has delivered and is contacting industrials stakeholders or organisations able to increase the visibility of SysCLAD and create new partnerships for the use of new biomarkers found or potential pathways to be targeted in view of novel therapies.
PROJECT AND INTELLECTUAL PROPERTY MANAGEMENT (CHUV/FINOVATIS)
Finovatis has been very instrumental for the implementation of a strong management structure to support the operational, technical and financial administration of the project. Identification of the best commercialisation strategy for the SysCLAD model in order to maximise technology transfer to clinicians and patients. The potential for a follow up of the SysCLAD cohort is being intensely scrutinized with the main partners of SysCLAD . Indeed many deliverables are promising to lead to new bridges between basic sciences discoveries and patient outcome management if all SysCLAD objectives could be implemented in the setting of system medicine.
List of Websites:
www.sysclad.eu