Final Report Summary - DECANBIO (Novel MS-based strategies to discover and evaluate cancer biomarkers in urine: application to diagnosis of bladder cancer)
Executive Summary:
Presently, thousands of genes or proteins can be characterized in one single experiment, providing new insights into biological processes. High throughput transcriptomics, and more recently, proteomics, enable comparisons of clinical samples from large populations of patients and controls. This represents a formidable opportunity for the discovery of early diagnostic or prognostic biomarkers, i.e. molecular indicators of a pathological state or its probable evolution that are present before the onset of specific clinical symptoms. The present challenge remains the integration of these technologies and all the data generated in a clinical set up in the discovery phase of novel biomarkers on one hand, and translating the findings into new clinical assays on the other. Through a four year collaborative program bringing together highly recognized experts in the fields of urology, clinical research, proteomics and transcriptomics, the DECanBio consortium has deployed considerable efforts aimed at the following objectives: (i) develop reliable analytical protocols and strategies enabling the generation of robust quantitative proteomics data, (ii) use them for the discovery and evaluation of new candidate biomarkers of bladder cancer in urine, and (iii) assess their predictive utility as surrogate endpoints through a rigorous pre-clinical evaluation.
To reach these ambitious objectives, the DECanBio project was subdivided in four consecutive phases:
A. A preparation phase during which the infrastructure for a large urine biobank was setup, and the analytical methodology elaborated, optimized and cross-validated;
B. A discovery phase devoted to the determination of lists of putative biomarkers of bladder cancer via mining of literature and web-based resources, and the study of tumor tissues and patient urines using a variety of experimental approaches;
C. An evaluation phase dedicated to the assessment of the detectability and discriminative power of the previously reported biomarker candidates in patient urines using dedicated mass spectrometry based methodologies developed within the project;
D. A validation phase in which the diagnostic performances of the verified candidates was evaluated using an orthogonal technology, namely micro-ELISA with electrochemical detection, also developed in the context of the project.
At the end of these four years, although not all objectives have been realized at 100%, the outcomes of the project are substantial and include:
1. A highly controlled biobank of urine samples from patients with a suspicion of bladder cancer (1,286 samples from 1,070 patients), including extensive clinical follow up information.
2. A standardized methodology for urine sample collection and subsequent protein extraction and digestion procedures dedicated to proteomic analysis.
3. An integrated pipeline for candidates biomarkers discovery and evaluation involving a high throughput unbiased discovery phase using Accurate Mass and Time tag strategy, and a highly sensitive targeted evaluation method using intelligent Selected Reaction Monitoring.
4. A urinary protein database containing 16,204 peptides corresponding to 1,936 non-redundant proteins, with information about their mass spectral signature.
5. Multiple lists of putative biomarker candidates:
a) Proteins from literature and internet resources mining
b) Genes differentially expressed in cancer from transcriptomics measurements on tumors.
c) Proteins differentially abundant in cancer from proteomics experiments.
6. A micro-ELISA system and methods with electro-chemical detection, including:
a) Microfluidic chips with integrated electrodes
b) Prototype of automated instrument for micro-immunoassays
c) Software for controlling electrochemical measurements in manual and automated instruments
d) Bead-coating and ELISA procedures for electrochemical ELISA tests using paramagnetic beads.
7. A panel of 24 urinary proteins verified to differentiate cancer patients from clinical controls.
8. A list of 21 bladder cancer specific splice variant genes, of which 7 protein products have been detected by SRM in urines from cancer patients.
9. Isotopically labeled QConCAT standards corresponding to:
a) The aforementioned 24 urinary protein marker panel.
b) The 7 aforementioned splice variant proteins.
10. The limited confirmation of the implication of a selection of candidates in bladder cancer and its evolution.
In conclusion, the work performed within DECanBio led to several highly significant outcomes. Among these, the determination of biomarker candidates for bladder cancer diagnosis and recurrence/progression constitutes a remarkable achievement. Provided it is followed up by appropriate validation efforts, these results could have major economic and sociological impact through: (i) decrease the number of the costly and highly uncomfortable and unnecessary cystoscopies; (ii) schedule in a more efficient way the surgical intervention (cystectomy) which severely compromises the quality of life of patients; (iii) guide the design of effective therapeutic interventions.
Project Context and Objectives:
Nowadays, thousands of genes, or the proteins they code for, can be analyzed in one single experiment, providing new insights into biological processes. These high throughput transcriptomics and proteomics measurements thus enable comparisons of clinical samples from large populations of patients and controls. This represents a formidable opportunity for the discovery of early diagnostic or prognostic biomarkers, i.e. molecular indicators of a pathological state or its probable evolution that are present before the onset of specific clinical symptoms. However, translating these fundamental discoveries into daily clinical practice is a long, tedious and costly process. In fact, in spite of intense efforts and investments, few novel biomarkers are used in clinical practice, and the rate of approved assays by the regulatory agencies is dismal. This is partly due to the lack of a linear and integrated process connecting candidate biomarker discovery to a high throughput, robust and sensitive platforms for their validation. Moreover, highly sensitive techniques are needed to overcome the tremendous complexity and the extremely broad dynamic range of protein concentrations encountered in bodily fluids proteomes. In other words, a major hurdle resides in the translation of clinical omics-technology into innovative biomarkers. Through a 4-year collaborative programme bringing together highly recognized experts in the fields of urology, clinical research, proteomics and transcriptomics, the DECanBio consortium has deployed considerable efforts in the context of bladder cancer to develop reliable quantitative analytical methods, and use them to identify new candidate biomarkers and begin to assess their predictive utility as surrogate endpoints.
To achieve these goals, the DECanBio work program was divided into four successive phases
1. Phase 1: Preparatory phase.
During this phase, the objectives were fourfold: (i) to set up the infrastructure for the collection of controlled cohort of patients with suspicion of bladder cancer, (ii) to optimize samples handling and preparation for proteomics experiments, (iii) to devise a comprehensive analytical strategy for biomarker discovery and evaluation by proteomic methods, and (iv) to start the development of a prototype of micro-ELISA system that could ultimately be used for biomarker validation.
Bladder cancer urine biobank:
The identification, selection and inclusion of the study subjects implied a very good coordination between clinical centers and the research teams to ensure both standardized conditions for sample collection as well as the inclusion of the desired number of individuals estimated in the project timeframe. At the beginning of the project, this required the development of a Study Protocol describing the details of the study design and includes questionnaires (M1.3) and standard operating procedures (M1.2) (D1.2). Its objective was to standardize methods among the participating centers and personnel involved in the study. Urine sample tracking forms were also needed to recapitulate all pertinent data required for sample and statistical analysis, including data related to sample processing and patient related data specific to the day of collection. These data were deemed necessary to provide both quality criteria for sample inclusion in the discovery step, and co-factors for statistical analysis. In addition, to satisfy legal requirements, the proper informed consent forms had to be compiled, and the protocol had to be submitted to each hospital's Ethics Committee for approval. To ensure high quality of the collected samples, it was projected that specific recommendations regarding their diet and exercise should be made to the patients prior to their appointment at the clinic. Finally, the project required the implementation of a common database dedicated to DeCanbio including a secure web-access (M1.4).
Critical issues with urine samples:
At the onset of the project, urinary proteomics being still in its initial stages, it was considered crucial to determine the optimum urine collection and sample preparation methods. Importantly, irreproducibility introduced at these early stages of the experiment would irremediably translate down the line into greater sample variability, potentially masking important signals, and eventually preventing discovery of relevant biomarkers. Thus, important efforts were anticipated to focus on a series of issues identified as critical in urinary proteomics: (1) addition of proteases inhibitors at collection stage, (2) urine storage temperature prior to centrifugation and freezing, (3) addition of preservatives to prevent bacterial growth, (4) influence of freeze-thaw cycles, (5) optimal protein extraction method and (6) minimization of technical variability.
Integrated strategy for biomarker discovery and evaluation.
Translating fundamental biomarker discoveries into clinical practice is a long, tedious and costly process. The major hurdle lies in the capacity to trim the often large lists of putative biomarkers emerging from high throughput discovery methods into more manageable sets of markers that have genuine clinical interest. This hurdle stems from the lack of a linear and integrated process connecting candidate biomarker discovery to a high throughput, robust and sensitive platforms for their evaluation. One major objective of the DECanBio project was the definition and implementation of such an integrated pipeline and its use in the context of a search for bladder cancer biomarkers.
Development of a micro-ELISA platform with electrochemical detection
Once putative biomarker lists have been pruned to the most interesting candidates, the second obstacle to clinical transfer resides in the ability to validate them in the clinical settings against large cohorts of patients. For this purpose, mass spectrometry has not yet gained general acceptance, and the more general route is to use ELISA tests. However, classical tests are time consuming and require large amounts of reagents. We thus projected to develop and thoroughly characterize a new highly automated micro-ELISA platform with electrochemical detection to facilitate the clinical validation process.
2. Phase 2: Biomarker candidates discovery phase
A wealth of information exists in the literature or on web-based resources concerning proteins involved in bladder cancer. The project plan involved curating this information and leveraging expert advice to prioritize a list of candidates from these external sources. In addition, since they have the capacity to compare samples from large populations of patients (healthy versus disease states), transcriptomics and proteomics open attractive avenues for new biomarker discovery. The initial discovery phase was designed to yield lists of gene products (RNA or proteins) displaying differential expression between normal and diseased samples based on semi-quantitative measurements.
Literature and web-based resources mining:
To leverage as much as possible of the prior knowledge about biomarkers for the system under study, we planned a thorough mining of available literature as well as web resources such as ProteinAtlas, a publicly available database with high-resolution images showing the spatial distribution of over 14,000 proteins in human tissues, cell lines, and cancer cells.
Bladder tumour transcriptomics data mining:
Genes significantly over-expressed in bladder tumors compared to normal urothelium code for proteins that are potential urine bladder cancer biomarkers. Using previously available transcriptomics data comparing normal urothelium and bladder tumour samples, the consortium was set to generate a list of gene products showing differential expression for subsequent evaluation using quantitative proteomics methods. (M2.1 D2.1). Further efforts aimed at defining alternative splice events in tumors were prompted by encouraging preliminary results that yielded a new avenue of research within the context of DECanBio.
Candidate biomarkers discovery by comparative proteomics:
Urine is readily available in large amounts using non-invasive collection procedures. Moreover, the urinary proteome contains a great variety of proteins. For bladder cancer, urine constitutes a proximal fluid, i.e. in direct contact with the diseased organ. As such, urine can be viewed as a local sink for proteins or peptides secreted, shed or leaked from the tumor. For this reason, urine represents highly relevant diagnostic analyte for bladder cancer detection using proteomics. Studying the urinary proteome was seen as a complementary approach to yield bladder cancer marker proteins that could not be predicted by transcriptome analysis of tumors. It was thus anticipated that new bladder cancer biomarkers could be derived through comparative analysis of the urinary proteome of healthy and diseased patients. In the Discovery Phase of DECanBio, protein abundance levels in patient urines had to be compared by two complementary proteomics approaches: quantitative two-dimensional gel electrophoresis (2DE) (D2.6) and Accurate Mass and Time tags proteomics (AMT) (D2.7 M2.4). Both technologies were selected to enable the unbiased discovery of bladder cancer biomarker candidates. An important aspect of these measurements was the availability of suitable urine samples. It was thus planned to provide urine samples from patients with a confirmed bladder cancer, and controls from donors, matched by age and sex, who did not present urological pathologies.
3. Phase 3: Biomarker candidates' evaluation phase
Because of the potentially high false positive rates associated with the breadth of the methods used at the candidate's discovery stage, the putative differentially abundant gene products, and their link to bladder cancer had to be carefully evaluated prior to forwarding them to the validation stage.
Method development:
To enable high throughput and reliable evaluation of the discovered candidates, Selected Reaction Monitoring (SRM) Mass Spectrometry methods had to be developed (M4.1 D4.3) to allow detection and accurate quantification of urinary proteins. In anticipation of the large number of transitions that would have to be followed, the development of acquisition scheduling was proposed. The objective was to enable the simultaneous evaluation of over 100 proteins in a single LC-SRM run.
Quantification standards:
For optimal performance of the quantitative measurements, quantification standards were required for all SRM targets included in the study. For this purpose, concatemers of tryptic peptides selected from each candidate protein sequence, called QConCAT standards, were developed and synthetized using a biotechnological approach. (M3.1 to M3.3) (D3.1 to D3.3)
Candidate qualification:
To assess the detectability and quantitative response of each candidate biomarker issued from the discovery phase, the plan called for submitting them to a first SRM screen (M4.2) that involved spiking synthetic standards into urine prior to analysis. These standards had to be designed to carry an isotopic label in order to differentiate them from endogenous biomarkers.
Candidate confirmation and verification:
Once the candidates had been qualified, the next projected step called for a confirmation of their differential abundance between the patients populations used at the proteomics discovery stage, i.e. bladder cancer patients vs. healthy controls. This had to be performed with adequate sample standardization.
Finally, the qualified candidates were submitted to verification of their predictive ability in differentiating urine samples from cancer patients and from patients with a suspicion of bladder cancer that had been cleared by pathological examination. Again, adequate sample standardization was required.
4. Phase 4: Biomarker candidate pre-clinical validation phase.
The last phase of the DECanBio project aimed at a pre-clinical validation of some of the findings about bladder cancer biomarkers using the technology developed by members of the consortium. This final goal was highly ambitious, and has unfortunately not been fully realized. As a matter of fact, delays accumulated throughout the two initial phases prevented the delivery of the list of verified biomarkers on time. In addition, benchmarking of the micro-ELISA technology yielded results that were below expectations in terms of reproducibility, which precluded its successful use in a clinical setting. The project management board, taking into account these factors, decided to entirely reorganise this final phase to allow preliminary validation of a selection of scientific results that were considered highly original and having strong potential for subsequent clinical applications. These included a set of candidate biomarkers having emerged from metallo-protein enrichment studies (D6.2 D6.3) and new potential bladder cancer detection methodology based on the study on cancer-specific alternate splice events.
Development and validation of micro-ELISA methods:
In spite of the concerns that had emerged from the technology benchmarking results, additional test were require to evaluate: (1) the analytical performance of micro-ELISA assays (including trueness, repeatability, and reproducibility), (2) their measurement range, and (3) the pre-analytical variation (including sample collection and handling, and the impact of physiological parameters (age, gender, weight, time of urine catch, etc.). For this purpose, urine samples of healthy controls and patients with bladder cancer from the discovery cohort were considered as analytes.
Pre-clinical validation of diagnostics biomarkers:
Among the proteins enriched through the metal-ion affinity fractionation strategy developed in the context of DECanBio, a set of proteins were considered of higher interest for further investigation, based on their functional annotation, and the lack of reported association with bladder cancer in earlier studies of urine.
Pre-clinical validation of prognostic biomarkers:
New findings that had emerged from the project relate to the existence of alternative splice isoforms that were found to be bladder cancer specific, and the possibility to detect these isoforms in cancer patients' urine. On the basis of transcriptomic data, the molecular taxonomy of bladder cancer could be reconsidered as suggested by Sjodahl et al. (A molecular taxonomy for urothelial carcinoma. Clin Cancer Res 2012).
Classification as Assay Specific Analytes:
It was initially planned that all biomarker assays, including MRM procedures, QconCATs and ELISAs, that would demonstrate their effectiveness during qualification, verification and/or validation steps, would be submitted for classification as Analyte Specific Reagents. Thus, provided accuracy and safety, they would be available for research use, while their potential clinical utility was being explored.
Project Results:
1. Phase 1: Preparatory phase:
a. Bladder cancer urine biobank:
The identification, selection and inclusion of the study subjects implied a very good coordination between clinical centers and research teams to ensure both standardized conditions for sample collection as well as the inclusion of the desired number of individuals estimated in the project timeframe. One hospital in France (H. Mondor in Creteil) and one in Spain (H. Mar in Barcelona) have been hired to cover: (i) patient identification selection and inclusion, (ii) data collection through both interviews and reviews of clinical charts, (iii) sample collection, and (iv) patient follow-up. The Study Protocol has been developed. This document described the details of the study design and includes the questionnaires and SOPs. Its objective was to standardize methods among the participating centers and personnel involved in the study. This study got the institutional review board (IRB) approval from H. Mondor and H. Mar. The Inform Consent forms have been designed and used in these hospitals.
Urine collection SOP: Taking into account preliminary results obtained by Partner P1 (CEA) in the context of WP2, recommendations for urine collection prepared in year 2007 through a proteomic fluid initiative coordinated by the French national institute of Cancer (INCa), and the clinical constraints; clinicians, epidemiologists, methodologists and proteomic researchers agreed on a Standard Operating Procedure. This SOP covers the volume collected, time of collection, delay elapsed between collection and freezing, storage temperature before freezing, protease inhibitors adjunction, and urine controls at bedside (hematuria, proteinuria, leucocytes, nitrites, diabetes) (M1.2). It was strictly followed by the urological centers involved in the study. Furthermore, a list of recommendation to the patients prior to collection and a questionnaire requesting information on the conditions of urine collection, have been developed. They have been translated in Spanish and French (M1.3). Also, questionnaires on epidemiological, clinical and follow-up data have been designed and translated to French from the Spanish version. A Web computerized data entry link to a database located on a CNIO server has been developed by the Bioinformatic Unit-INB at CNIO, Spain (M1.4).
Collection Plan: The collection plan for the various cohorts was finalized during the first period and periodically reevaluated. The project requirements called for three cohorts, the first for the candidate discovery studies (D1.3) the second for evaluation purposes (D1.4) and the last for the validation effort (D1.5). For the discovery phase, the definition of the population to be analyzed (90 samples) and discussions related to clinical organization (including the rate of patients actually diagnosed in the participating centers) allowed to plan the collection of the initial 90 samples that were used by Partners P1 (CEA) and P2 (BRFAA) during the discovery phase. To ensure the achievement of 60 urines samples in patient with bladder cancer (30 incident and 30 prevalent cases), it was estimated necessary to include at least 600 patients submitted to cystoscopy.
b. Critical issues with urine samples:
Urine proteome does not present a stable pattern of proteins; it varies depending on the sampling procedure (time, conditions, storage and handling) as well as the age, gender, nutrition, presence of disease, drug administration of the donor. Despite these inherent difficulties in the analysis of the urinary proteome, protein markers for diseases (including bladder cancer) have been detected in urine and have been approved to be utilized as adjuncts for disease diagnosis and prognosis. Therefore, an important task of DECanBio partners during the first period was to develop standardized approaches for urine sample handling, storage and processing (D1.2 D2.5).
In a clinical setting, urine collection may involve spiking the sample with proteases inhibitors and cold storage (0-4°C) for an extended period of time (up to 6 hours) before low speed centrifugation to eliminate cellular debris. However, considering that proteases have ample time to operate at 37°C during urine retention in the bladder, it was questionable whether adding inhibitors after collection resulted in significant changes in the urinary proteome. Furthermore, knowing that many proteins (e.g. uromodulin) do precipitate at cold temperature under salt conditions, the storage temperature before centrifugation had to be tested also. Therefore, in the context of DECanBio, a dedicated series of experiments have been design to investigate the use of protease inhibitors and the storage temperature. Several metrics were used to evaluate the influence of these pre-analytical conditions: (1) the total number of peptides identified per condition; (2) the median abundance of detected peptides per condition; and (3) the correlation coefficient between peptide abundances measured in treated samples vs. controls. All the results converged to the conclusion that the most favorable conditions for urine storage after collection were to add proteases inhibitors immediately after collection, avoid cooling the sample and, most importantly, to process them in a timely manner.
c. Integrated strategy for biomarkers discovery and evaluation.
As pointed out by Rifai et al. (Nature Biotechnology 2006, 24: 971-983), a major hurdle in putting biomarkers to clinical use lies in our incapacity to trim the extensive lists of putative biomarkers emerging from high throughput discovery studies into smaller sets of markers that have genuine clinical interest. This hurdle stems from the lack of a linear and integrated process connecting candidate biomarker discovery to a high throughput, robust and sensitive platforms for their evaluation.
d. Development of a micro-ELISA platform with electrochemical detection
One long term objective of DECanBio was to develop a microfluidic ELISA test (μELISA) system adapted to the detection of a few verified DECanBio cancer biomarkers in urine samples. Microfluidic ELISAs would offer many advantages over conventional microtiter plate assays in routine analysis in the doctor office since they will use less reagent material and provide more rapid testing. To this end, the DiagnoSwiss Company has adapted its original electrochemical microsystems by producing new batches of microchannel arrays and by coupling them with magnetic nanoparticles. One key element of this system was to enable to process samples of large volume (namely up to a few hundreds of μL) compared to the few hundreds of nL of the microchannel, and hence to pre-concentrate biomarkers on the magnetic nanoparticles, thereby increasing the sensitivity of the system. During the second reporting period of DECanBio the development of the microfluidic chip platform was pursued. This has led to the development of the ImmuSpeed, a prototype of industrial instrument (D5.4) which includes a pipetting unit, four positions for plates, two shakers, one washing station and one module for beads pre-concentration and detection in DiagnoSwiss microchip consumables. The system further features an internal personal computer and its proprietary software that enables to create assay protocols, to control the operations of the instrument and the multiplexed electrochemical detector and to process the row measurement data. The ImmuSpeed system has been dimensioned so as to have a capacity of up to 96 data points processed in series of 8 parallel analyses from standard and sample solutions embarked in a microtiterplate. The reagents are placed in the other plate positions of the machine and the microchip is regenerated between two series of experiments.
In conclusion, if the micro-ELISA approach showed good performances in terms of sensitivity, detection time and reagent volume, the developed platform was still at a prototyping stage, and the results obtained in urine samples were not robust enough to enable transfer to clinical biological practice (lack of reproducibility and difficulty in handling magnetic beads and urine samples). For these reasons, the prototype could not be used for the validation phase as initially planned, which prompted the development of an alternative plan for the final phase of the project.
2. Phase 2: Biomarker candidates discovery phase
Recent investigations have demonstrated that the urinary proteome contains a much greater variety of proteins than previously recognized. For bladder cancer, in contrast to blood, urine is a proximal fluid, i.e. directly in contact with the diseased organ. As such, urine can be viewed as a local sink for proteins or peptides secreted, shed or leaked from the tumor. In addition, urine is readily available in large amounts using non-invasive collection procedures. For all these reasons, urine represents an ideal diagnostic analyte for bladder cancer detection using proteomics strategies. The Biomarker Candidates Discovery phase of DECanBio consisted in a multi-pronged search for potential biomarkers of bladder cancer. For this purpose, literature and web resources mining, transcriptomics, and proteomics strategies were put to use to try and come up with consistent sets of proteins that should be considered for further evaluation.
a. Literature and web-based resources mining.
The DECanBio fits in a broader context of bladder cancer research, and a wealth of data had previously been accumulated on this topic in the literature using a variety of experimental approaches. The project plan called for a thorough evaluation of existing resources to select and prioritize sets of proteins previously considered characteristic of bladder cancer in the hope of detecting them in patients' urines.
b. Bladder tumour transcriptomics data mining
Genes significantly over-expressed in bladder tumor samples compared to normal urothelium code for proteins that the tumor might shed in the urine, and thus constitute potential bladder cancer biomarkers. A set of transcriptome data was available to DECanBio partners: it was obtained by Partner P4 (IC) and Partner P6 (AP-HP) in collaboration with AstraZeneca. In this series of data, the transcriptome of normal tissues and had been studied on the Affymetrix U95A array. One of DECanBio objectives was to assess these existing data using dedicated statistical methods taking into account tumour heterogeneity. This was expected to lead to the identification of roughly 200 differentially expressed genes. We first searched for the 300 most significantly overexpressed in bladder tumors (n=80) compared to normal urothelium (n=5) using the SAM software. We then examined visually all the 300 profiles. Strikingly, many of the genes overexpressed in tumors were also overexpressed in the chorion or the smooth muscle, suggesting that the overexpression observed in tumors could be not due to the transformed cells but to the stroma of the tumor. We decided to retain only genes which were significantly overexpressed in tumors compared to normal urothelium and also overexpressed in more than 25% of tumors compared to the chorion and to the smooth muscle. In addition we have also added a few genes which were obtained by specifically examining the expression profiles of family of genes which could play a role in cancer (Interleukins, CXCLs, IGFBPs, IGF). Among these genes, those which were expressed in any of the normal or inflammatory tissues were selected as new bladder cancer biomarker candidates. This approach led to a total of 80 genes whose products were considered for further evaluation (D2.1).
To complement this work, Partner P4 (IC) proposed to the Project Management Board to consider a new list of bladder cancer biomarker candidates that would correspond to splice variants specifically expressed in bladder tumors. This was approved as an extension to the initial DECanBio workplan. As a matter of fact, Partner P4 had started to analyze the expression data of bladder tumors, normal urothelium, chorion and smooth muscle on an Affymetrix exon 1.0 array. In classical chips, when they are several transcripts per gene, the measured expression is the total expression of all the isoforms as the probes are in general chosen to be common to all isoforms. In the exon arrays, as they are probes for most of the exons of the gene, one can measure individually the exons and therefore can try to infer splicing events. The idea was to find splicing events which 1) were tumor specific (not find in the normal samples), 2) would lead to the addition of new amino acids in the proteins or deletion of amino acids. Both addition and deletion would change the sequence of the proteins and therefore should lead to 1) new peptides after protease digestion which could be detected by mass spectrometry; and 2) create new epitopes which could be detected using antibodies. This work resulted in a list of 55 new peptides that should appear in the proteins as a result of these splicing events. In addition, a patent was submitted by Partners P4 (IC SR) and P6 (AP-HP) in December 2011, both on the markers obtained based on classical arrays and the markers obtained using the exon arrays (Urinary markers for monitoring bladder cancer).
c. Candidate biomarkers discovery by comparative proteomics
Looking directly at the urine proteome is a complementary approach that could be expected to lead to the identification of bladder cancer marker proteins that could not be predicted by transcriptome analysis of tumors. Therefore, new bladder cancer biomarkers were expected from the thorough analysis of the urinary proteome of healthy and pathological conditions. In the discovery phase of DECanBio, protein abundance levels have been determined by quantitative two-dimensional gel electrophoresis (2DE) and by using accurate mass and time (AMT) tag proteomics approach. Both technologies enabled the unbiased discovery of bladder cancer biomarker candidates.
i. 2DE-based discovery
The first methodology employed in DECanBio for the identification of potential bladder cancer biomarkers in urine was based on the application of comprehensive electrophoretic approaches. In brief, preparative electrophoresis in combination to 2-dimensional electrophoresis was used for the efficient and high resolution separation of urinary proteins.
Preliminary work focused on down-scaling and optimizing the prep-2DE technique so as to make it applicable for the analysis of urine samples from bladder cancer patients. Our specific aim was to optimize the various experimental steps involved in the workflow including desalting/purification of protein fractions, conditions for 2DE analysis as well as detection system, so as to avoid, the potential decrease in protein resolution imposed by down-scaling.
The subsequent step consisted in the development of a gel-based urinary proteome database. This work was performed in collaboration with the EuroKUP (European Kidney and Urine Proteomics COST Action) consortium. Specifically, two normal urine samples corresponding respectively to pooled samples from male and female healthy volunteers, were collected within EuroKUP in large quantities and distributed to multiple laboratories with the objectives to be characterized comprehensively by multiple state-of the art proteomics methods and be used as standards in clinical proteomics applications. Partner P2 (BRFAA) generated the 2DE proteomic map of this sample which included in several cases information on protein post-translational modifications. In addition a side-by side comparison of the 2DE and shotgun data collected by Partner P1 (CEA) was conducted, clearly revealing the individual advantages and complementarities of the different techniques.The results of this multi-institutional analysis have been published in the journal Proteomics Clinical Applications.
ii. AMT tag discovery
Partner P1 (CEA) has long been involved in the development and implementation of the Accurate Mass and Time (AMT) tag methodology, a high throughput quantitative proteomics strategy. The first phase of this strategy involves compiling a so-called AMT tag database of identified peptides and their parent urinary proteins from patient's urine samples having been subject to various biochemical pre-fractionations. The second phase allows the comparison of peptide abundances between large numbers of samples, as required for the comparison of patient cohorts. AMT tags proteomics thus appears very promising for candidate biomarkers discovery, insofar as it facilitates high-throughput extensive proteome analysis. This strategy was used in the context of DECanBio for label-free comparative proteomics study of urine from bladder cancer patients vs. controls.
An AMT tag database has been compiled throughout the DECanBio project (D2.4) using a variety of urine samples from diverse origins. These samples included healthy donors, patients with suspicion of bladder cancer that was subsequently cleared, and patients with non-invasive bladder cancer corresponding to various risk factors. This task required the implementation of a variety of sample fractionation techniques such as the Proteominer, and a large analytical effort. The final version of this database contained data acquired in over 1,200 LC-MS/MS analyses of urine samples. This database constitutes an in-depth repository of all peptides previously detected in urine by our group and their corresponding proteins. In addition, it contains information about peptides MS/MS spectra, which could be queried to determine lists of transitions to monitor during the biomarker evaluation phase. The final version of our AMT tag database contained over 18,000 non-redundant peptide entries, representing approximately 2,000 urinary proteins (D2.4). This AMT database was subsequently used in the DECanBio project to identify candidate biomarkers for bladder cancer, but it could also be put to use for any disease in which analysis of the urinary proteome might be relevant, e.g. to identify markers of kidney disease. Implementation of generic tools was one of the objectives of DECanBio; the AMT database of the urinary proteome can be seen as one such tool. It is publicly available on the DECanBio website (http://www.decanbio.eu).
3. Phase 3: Biomarker candidates' evaluation phase
As pointed out earlier, the breadth of the methods used at the candidate's discovery stage was associated with potentially high false positive rates associated. This was unavoidable at this stage due to the low statistical power associated with the multidimensional measurements required for the wide screening process. Therefore, the putative differentially abundant gene products, and their link to bladder cancer had to be carefully evaluated prior to forwarding them to the validation stage. One of the critical objectives of DECanBio was thus the thorough evaluation using SRM assays of three series of biomarkers candidates: the first originating from literature and web-resources mining, the second from proteomics measurements, and the third from recently obtained exon-chip transcriptomic assays. This evaluation was setup in three consecutive steps: (1) candidate qualification, consisting in checking the detectability and the linearity of the quantitative response of the corresponding peptides in urine; (2) candidate confirmation, i.e. assessment of the differential concentration of a candidate between urine from healthy donors and urine from cancer patients, thereby confirming, when available, the results obtained during the discovery phase; (3) candidate verification, involving the accurate estimation of the differential concentration of the candidates in urines from cancer patients vs. urines from clinical controls having a suspicion of bladder cancer and for whom the suspicion had been cleared by pathological examination. Before this work could be undertaken, significant efforts had to be devoted to the development and optimization of methods and standards to achieve the evaluation task.
a. Method developments.
A proteomics screening experiment is performed at peptides level, and the selection of targets uniquely associated to the protein of interest. In contrast to conventional shotgun proteomic studies, selected reaction monitoring (SRM) measurements are quantitative, rigorously targeting a predetermined set of peptides. The determination of the proper SRM transitions for each target is an essential step in biomarker evaluation. More precisely, for each protein of interest, the peptides presenting good ionization properties and which are unique surrogate of the protein (or a specific isoform), called proteotypic peptides (PTPs), need to be selected. For each peptide, fragment ions that exhibit strong signals in MS/MS spectra and discriminate the targeted peptide from other species were preferred. In addition, other parameters critical to SRM analysis, such as collision energy also required optimization.
b. Quantification standards
As initially evocated in the technical risks assessment of DECanBio DoW, the protein content of urine samples varies depending on the age, sex, diet, medication, time of sample collection etc. All these factors may complicate the interpretation of results when dealing with biomarker discovery. One way to take these factors into consideration is the adjustment of the protein concentration of the urine samples under comparison. This was achieved by protein concentration measurement using pyrogallol red and Bradford assays. In addition, for absolute quantification of urinary proteins, the adjunction of quantification standards was a necessary requirement. We opted for QconCAT standards, concatemers of tryptic peptides selected from each candidate protein sequence.
The project thus required the synthesis of multiple QconCAT polypeptides covering a list of potential biomarkers for the diagnosis of incident and recurrent bladder cancer against a background of healthy subjects and patients with other, non-cancerous bladder-related complications. Each QconCAT consists of multiple isotope-labeled proteotypic peptides. Each QconCAT could be spiked into a biological sample, and after digestion with trypsin, the contained reference peptides were released, as well as their counterpart peptides from the natural proteins present in the sample. By comparing peak intensities in a mass spectrum of natural and reference peptides, the original amount of native protein present in the sample could be determined for all proteins covered by a QconCAT.
Partner P7 (Entelechon) was in charge of the development and production of isotopically labelled standard polypeptides (QconCAT's) to be used by Partner P10 (CRP-Sante) for the detection and accurate quantification of bladder cancer biomarker candidates in urine samples by mass spectrometry. This included the bioinformatic optimisation of the expression constructs for the concatamerisation of the selected putative proteotypic peptides, the gene synthesis of these constructs, as well as the cloning, expression and labeling of the QconCAT standards.
c. Candidates qualification
A critical part of biomarker candidate evaluation consisted in the development and the implementation of a workflow to translate a protein/transcript list of candidates into peptides used to perform a wide screen using the MS-based analysis in urine. It had a first objective to detect and identify the peptides (and per association the proteins) in urine samples. This was called the biomarker candidate qualification phase.
The objective of this work was to confirm the detectability in urine samples of the proteins from three lists of prioritized bladder cancer biomarker candidates: The first priority list, called Set 1 containing 152 protein candidates, had been selected based on their biological and clinical relevance. The selection of these candidates was a compilation of results from the scientific literature, combined with protein, for which a bladder cancer specific expression profile was observed based on immunohistochemistry patterns for a large number of human tissues, cancers and cell lines, available through Human Protein Atlas (www.proteinatlas.org). In addition, 80 proteins corresponding to genes, whose expression was shown to be dysregulated in tumors by Partner P4 (IC-SR) were appended to this list. A second set of biomarker candidates, called Set 2 encompassing 55 proteins, was composed of proteins emerging from two biomarker discovery studies (Task 2.3 and 2.4) performed by Partners P1 (CEA) and P2 (BRFAA). Finally, a third set, Set 3, was obtained from transcriptomic data, and with a focus on skipped/added exon data by Partner P4 (IC SR). Select proteins from these three sets of candidates were subsequently evaluated using robust instrument methods for SRM analysis developed previously.
Protein Candidates Set 1: A series of 33 proteins (Top 33) resulting from the confrontation of Set 1 with the urinary proteome AMT database (D2.4) created by partner P1 (CEA) within, formed the basis for a primary SRM screen. Eighty six peptides, surrogates of the 33 protein candidates, were unambiguously detected in patient urines and their identity confirmed using stable isotope labeled synthetic peptides. A second set of 41 proteins from the 152 entries priority list of candidates had been selected and screened. Sixty eight peptides, surrogates of 37 protein candidates, were unambiguously detected and confirmed with the stable isotope labeled peptides, while four proteins were not detected by monitoring the corresponding peptides. Three additional proteins from Set1 and eight other new proteins candidates were selected based on their clinical relevance and new evidence gathered from studies that became available within the project timeframe (literature mining, transcriptomics data). In total, 82 out of 152 protein candidates from Set 1 were qualified during this stage of the study (D4.4). The results of this screening experiment led to the design and synthesis of four isotopically labeled QconCAT standards by Partner P7 (Entelechon).
Protein Candidates Set 2: The screening of 55 protein candidates from the proteomics Discovery phase of DECanBio performed by Partner P1 (CEA) and P2 (BRFAA) in the context of the discovery phase was performed. From these 55 candidates, 17 were already screened in Set 1. One hundred seven peptides, surrogates of the remaining 38 protein candidates, were unambiguously detected and confirmed using stable isotope labeled peptides. The results of these experiments have led to the design of two additional isotopically labeled QconCATs standards that were synthetized by Partner P7 (Entelechon).
Protein Candidates Set 3: The candidates in Set 3 were constituted by 19 gene chip exons related to 21 gene candidates. The amino acid sequences (protein level) resulting from the translation of the gene exon candidates (mRNA), termed proteo-exon were selected. For each proteo-exon, a tryptic sequence covering the variable region (added or skipped sequence string) was selected to be monitored, together with five unique canonical peptides per protein. From the 19 original gene chip exons corresponding to 21 gene candidates, 9 proteo-exons related to 8 gene candidates were selected in accordance with SRM detectability criteria (e.g. amino acid composition, hydrophobicity, sequence length, etc). Upon data evaluation, two cases were observed: i) the detection of the proteo-exon and the canonical proteotypic peptides from the same protein candidate and ii) the sole detection of the proteo-exon, without the corresponding canonical sequences, suggesting the existence of other modifications on the corresponding gene products. The outcome of this screening resulted in the design of a QconCAT standard that was commissioned to Entelechon (FP7 Partner).
d. Candidate confirmation and verification
The subsequent phase of the work was to perform the quantification of the 134 qualified biomarker candidates in clinical urine samples from the so-called verification cohort (D1.4) in order to evaluate the differential abundance between physiological and pathological concentration, using high throughput SRM methodologies that were setup previously.
One hundred twenty three samples provided by partners P6 (AP-HP) and P9 (CNIO) were received and processed (peptide level) using the standardized protocol established by partners P1 (CEA) and P3 (ETH). This collection included urine samples from few healthy individuals, from patients with urological disorders for whom the cancer diagnosis had been cleared (real-life controls), and from bladder cancer patients stratified according to their EUA risk group (low, intermediate, and high), with associated demographic and clinical information. Four categories of patients could thus be distinguished: Incident: Patients with a first confirmed incidence of Bladder cancer; Incident control: Patients with no history of bladder cancer diagnosed with other urological diseases, Prevalent: Patient with a prior history of bladder cancer who, during their clinical examination, were diagnosed with bladder cancer recurrence; Prevalent control: Patient with a prior history of bladder cancer, whose examination results cleared any suspicion of bladder cancer recurrence.
4. Phase 4: Biomarker candidate pre-clinical validation phase.
The pre-clinical validation phase was originally planned as a way to transfer the list of urine bladder cancer biomarkers emerging from the discovery and evaluation phases to the clinic through immunoassays validated on independent cohorts, with clinical data and urine samples prospectively collected in the context of the project. During the DeCanBio project, considering recent development of SRM technologies, and the difficulties in obtaining immunoassays of good reproducibility in urine, the management committee of DeCanBio decided to perform the pre-clinical study of candidate biomarker through SRM. In addition, it appeared later in the project that the use of specific transcript splicing events could be a new source of biomarkers. As recently analysis of RNA extracted from urine has been shown to be feasible on a large scale, the consortium therefore started to develop assays to directly measure these splice forms in the urine.
During the project s third period, it became acutely evident that the rescheduling of series of tasks (onset of collection delayed due to changes in Spanish legislation, subsequent postponing of cohorts deliverables, rescheduling of the candidate discovery phase) that had accumulated since the beginning of the project could not be expected to be recovered during this final period. . In addition, the list of candidate biomarkers being still at the evaluation stage, it was not possible yet to proceed to their pre-clinical validation. The project management board thus decided to explore alternative ways to validate other findings obtained in the context of the project instead of the initially planned pre-clinical validation of the verified candidates. As a matter of fact, several unanticipated results, that could have significant impact on the understanding and/or diagnosis of bladder cancer, required additional proof.
a. Preliminary validation of candidates emerging from the metallo-protein screen.
Among the proteins enriched through the IMAC fractionation strategy developed by Partner P2 (BRFAA) during the preparation phase of DECanBio, aminopeptidase N, myeloblastin and profilin 1 were considered of higher interest and therefore prioritized for further investigation, based on their functional annotation, and the lack of reported association with bladder cancer in earlier studies of urine. In addition, during period, IMAC fractionation was combined with capillary electrophoresis- mass spectrometry (CE-MS; in collaboration with Dr H Mischak) for the study of the low molecular weight proteome (peptidome). This study was an expansion of an existing collaboration applying CE-MS to identify urinary peptides associated with bladder cancer staging. This study identified four urinary peptides, derived from uromodulin, collagen type I and III, and membrane-associated progesterone receptor, that enabled, in combination with cancer grade, the separation of muscle-invasive bladder cancer from other cancers, with had shown a negative predictive value of 77% and a positive predictive value of 90% in a previous blinded evaluation (Schiffer and Vlahou et al., Clin Cancer Res 2009).
b. Preliminary validation of splice variant results
The project management board decided to redirect validation efforts toward the assessment of some highly innovative findings made during the final period. These findings relate to the existence of alternative splice isoforms that were found to be bladder cancer specific, and the possibility to detect these isoforms in cancer patients' urine.
As a first step, Partner 6 (AP-HP) performed systematic immuno-histochemical studies using antibodies targeting some candidates potentially associated with the molecular taxonomy, among them cytokeratin 18, cytokeratin 20, Topo2a, PI3, FoxA1, S100A7, S100A8. Protein expression was assessed using tissue microarrays gathering the cases previously studied at transcriptomic level. For most of the candidates, the correlation between mRNA and protein expression was good or excellent. Preliminary analysis according to the stage and the grade showed frequent association with the expression. Further investigation will aim to propose a combination of biomarkers covering all the sub-phenotypes observed in bladder cancer, in order to test them in urine and estimate their positive and negative predictive values.
Potential Impact:
According to current clinical practice, the standard care for the detection of bladder cancer is cystoscopy and pathologic examination. Thus, urine cytology remains, despite its limitations, the standard non-invasive method for detecting bladder cancer, keeping a superior specificity to most available tumor markers. For this reason, no tumor biomarker could be recommended for use in bladder cancer screening at the onset of the project. Thus, the development of biomarkers for bladder cancer was still a matter of research. DECanBio implemented a novel workflow and unique state-of-the-art technologies: (i) to discover biomarker candidates by integrating various omics technologies; (ii) to qualify and verify these candidates using SRM-LC-MS technique which allowed to rapidly screen hundreds of candidates and to establish a reduced panel of biomarkers that could be the starting point of future validation efforts using dedicated SRM or ELISA assays. The most practical use of non-invasive tests could be directed to reduce the number of surveillance cystoscopies performed during the monitoring of patients treated for bladder tumors, and to detect precociously bladder tumor progression and dissemination by setting clinical protocols with frequent and convenient urine tests. The results of DECanBio research could indirectly affect not only the population suffering from bladder cancer but could have a major impact on the diagnostics/prognostics of multiple diseases such as the very prevalent kidney diseases, prostate cancer and others.
The scientific impact of DECanBio in terms of translational research has been very prominent as evidenced by the feedback received during the numerous presentations of the consortium's efforts in this area. In particular, the emergence of an integrated method for biomarkers discovery/evaluation based on the AMT tag methodology for the discovery phase and the iSRM strategy for the evaluation phase can be considered highly significant.
One of the critical objectives of DECanBio was the thorough discovery of urinary proteins with a potential link to bladder cancer. The corresponding discovery phase culminated in the delivery of three series of biomarkers candidates: the first originating from literature and web-resources mining, the second from proteomics measurements, and the third from recently obtained exon-chip transcriptomic assays. While a significant portion of these candidates has been subject to evaluation, we necessarily had to make a prioritization and not all candidates could be considered within the context of the project. These lists contain relevant targets for future evaluation efforts, and the consortium will have to decide how to make them available to the public for future studies.
An important task of DECanBio partners during was to develop standardized approaches for urine sample handling, storage and processing. Thus, important efforts focused on a series of issues identified as critical in urinary proteomics: (1) Addition of proteases inhibitors at collection stage (2) Urine storage temperature prior to centrifugation and freezing. (3) Addition of preservatives to prevent bacterial growth. (4) Influence of freeze-thaw cycles (5) Optimal protein extraction method (6) Minimization of technical variability. Standards operating procedures have been developed and published in the scientific literature, helping set guidelines for future urinary proteomes studies, not only in the context of bladder cancer, but for a variety of other pathologies.
A highly significant outcome of the DECanBio was obtained at the candidate evaluation phase, yielding a panel of 24 urinary proteins distinguishing bladder cancer patients from urologic controls. This panel constitutes the starting point for a future bladder cancer test, providing adequate validation efforts are undertaken. Mandate will be granted to Partner P7 (Entelechon), in consultation with the project coordinator and other members who have expressed their interest, to approach a diagnostic company interested in building a clinical diagnostics assay on the obtained results.
Another very important result that emerged from DECanBio project relates to the new concept for cancer detection involving spliced exons. The measurement of splicing events instead of global expression level of genes could significantly improve the specificity and the sensitivity of bladder cancer detection either for the initial diagnosis (patients presenting with hematuria) or for the follow-up. Furthermore, upon the development of ELISA or SRM assays specific of the spliced sequences, the splicing variants will also become measurable at the protein level as evidenced by their detection during the biomarkers evaluation phase of DECanBio. A patent was filed by Partners P4 (IC SR) and P6 (AP-HP) on this novel concept.
An important exploitable foreground generated in the frame of the DeCanBio project refers to the work of Partner P8 (Diagnoswiss) on the newly developed electrochemical microfluidic chips, the prototype of robotic analysis platform and specific test protocols for bead-based immunoassay applications. This foreground aims at obtaining a finalized first-series instrument applicable for beta-testing in view of finding a strategic partnership for the commercialization of a novel automated Elisa instrument or the transfer of the technology for R&D applications. The development of microfluidic systems as well as novel approaches reducing the number of assay steps (and hence of manipulations) opens new routes for the application of immunoassays in research laboratories and medical diagnostics, with the benefits of reducing the analysis time and cost. The results of the DeCanBio project confirm that the microfluidic chips, the robotized infrastructure and the developed assay protocols enable to perform unattended bead-based immunoassays with electrochemical detection that can be applied to multiple sequential analysis of specific proteins or biomarkers. In addition to the automation of the assays, the main advantages of the developed micro-ELISA technology rely on the reduction of the reagent consumption (and hence of the unitary test cost) and on the shortening of the time-to-result compared to traditional microtiter-plate methods. Application to urine samples has yet shown limitations due to the complexity of the sample matrix which interferes with the formation of the immune complex or induces strong background masking the effectively measurable signals. If the assay development work shows the flexibility of the developed electrochemical microchip platform in terms of applicable assay types and analyte concentration range, more relevant biochemical reagents are required to improve the selectivity and reproducibility of the approach with urine samples. Otherwise, further expertise will still be required to improve the handling of the magnetic beads and of the very low volumes involved at each assay step, and to better assess the robustness and long-term use of the platform in real conditions. Such additional work is envisaged for transforming the foreground generated in this project into a final product adapted to routine analysis in R&D or medical diagnostic laboratories.
In terms of direct social impact, DECanBio has resulted in the hiring of ten personnel (five women and five man) in positions entirely dedicated to the project. They included among others:
- A technician biochemist (Ms Madalen LeGorrec) and a computer scientist (Ms Claire Adam) by Partner P1 (CEA) to perform sample preparation and AMT tag database compilation and management respectively.
- An engineer statistician (Mr Mourad Mellal) and a post-doctoral associate (Dr Aurelie Kamoun) by Partner P4 (IC SR) to perform statistical analyses of AMT tag data and splice variant data respectively.
- A post-doctoral associate (Dr Elodie Duriez) for SRM analyses by Partner P10 (CRP-Sante).
- An assistant engineer (Mrs Edwige Lopez-Perreira) to supervise urine samples collection at Partner P6 (AP-HP).
In terms of advancing higher education, the project has been central to the PhD thesis of Mrs Magali Court and the Diploma work of Mr Mourad Mellal at Partner P1 institution (CEA). The PhD thesis topic aimed at the characterization of the urinary proteome, and of its potential for disease biomarkers research. The diploma work focused on the use of penalized logistic regression for the interpretation of comparative proteomics data.
In conclusion, the work performed within DECanBio led to several highly important outcomes. Among these, the determination of biomarker candidates for bladder cancer diagnosis and recurrence/progression constitutes a remarkable achievement. Provided it is followed up by appropriate validation efforts, these results could have major economic and sociological impact through: (i) decrease the number of the costly and highly uncomfortable and unnecessary cystoscopies; (ii) schedule in a more efficient way the surgical intervention (cystectomy) which severely compromises the quality of life of patients; (iii) guide the design of effective therapeutic interventions.
List of Websites:
http://www.decanbio.eu
Presently, thousands of genes or proteins can be characterized in one single experiment, providing new insights into biological processes. High throughput transcriptomics, and more recently, proteomics, enable comparisons of clinical samples from large populations of patients and controls. This represents a formidable opportunity for the discovery of early diagnostic or prognostic biomarkers, i.e. molecular indicators of a pathological state or its probable evolution that are present before the onset of specific clinical symptoms. The present challenge remains the integration of these technologies and all the data generated in a clinical set up in the discovery phase of novel biomarkers on one hand, and translating the findings into new clinical assays on the other. Through a four year collaborative program bringing together highly recognized experts in the fields of urology, clinical research, proteomics and transcriptomics, the DECanBio consortium has deployed considerable efforts aimed at the following objectives: (i) develop reliable analytical protocols and strategies enabling the generation of robust quantitative proteomics data, (ii) use them for the discovery and evaluation of new candidate biomarkers of bladder cancer in urine, and (iii) assess their predictive utility as surrogate endpoints through a rigorous pre-clinical evaluation.
To reach these ambitious objectives, the DECanBio project was subdivided in four consecutive phases:
A. A preparation phase during which the infrastructure for a large urine biobank was setup, and the analytical methodology elaborated, optimized and cross-validated;
B. A discovery phase devoted to the determination of lists of putative biomarkers of bladder cancer via mining of literature and web-based resources, and the study of tumor tissues and patient urines using a variety of experimental approaches;
C. An evaluation phase dedicated to the assessment of the detectability and discriminative power of the previously reported biomarker candidates in patient urines using dedicated mass spectrometry based methodologies developed within the project;
D. A validation phase in which the diagnostic performances of the verified candidates was evaluated using an orthogonal technology, namely micro-ELISA with electrochemical detection, also developed in the context of the project.
At the end of these four years, although not all objectives have been realized at 100%, the outcomes of the project are substantial and include:
1. A highly controlled biobank of urine samples from patients with a suspicion of bladder cancer (1,286 samples from 1,070 patients), including extensive clinical follow up information.
2. A standardized methodology for urine sample collection and subsequent protein extraction and digestion procedures dedicated to proteomic analysis.
3. An integrated pipeline for candidates biomarkers discovery and evaluation involving a high throughput unbiased discovery phase using Accurate Mass and Time tag strategy, and a highly sensitive targeted evaluation method using intelligent Selected Reaction Monitoring.
4. A urinary protein database containing 16,204 peptides corresponding to 1,936 non-redundant proteins, with information about their mass spectral signature.
5. Multiple lists of putative biomarker candidates:
a) Proteins from literature and internet resources mining
b) Genes differentially expressed in cancer from transcriptomics measurements on tumors.
c) Proteins differentially abundant in cancer from proteomics experiments.
6. A micro-ELISA system and methods with electro-chemical detection, including:
a) Microfluidic chips with integrated electrodes
b) Prototype of automated instrument for micro-immunoassays
c) Software for controlling electrochemical measurements in manual and automated instruments
d) Bead-coating and ELISA procedures for electrochemical ELISA tests using paramagnetic beads.
7. A panel of 24 urinary proteins verified to differentiate cancer patients from clinical controls.
8. A list of 21 bladder cancer specific splice variant genes, of which 7 protein products have been detected by SRM in urines from cancer patients.
9. Isotopically labeled QConCAT standards corresponding to:
a) The aforementioned 24 urinary protein marker panel.
b) The 7 aforementioned splice variant proteins.
10. The limited confirmation of the implication of a selection of candidates in bladder cancer and its evolution.
In conclusion, the work performed within DECanBio led to several highly significant outcomes. Among these, the determination of biomarker candidates for bladder cancer diagnosis and recurrence/progression constitutes a remarkable achievement. Provided it is followed up by appropriate validation efforts, these results could have major economic and sociological impact through: (i) decrease the number of the costly and highly uncomfortable and unnecessary cystoscopies; (ii) schedule in a more efficient way the surgical intervention (cystectomy) which severely compromises the quality of life of patients; (iii) guide the design of effective therapeutic interventions.
Project Context and Objectives:
Nowadays, thousands of genes, or the proteins they code for, can be analyzed in one single experiment, providing new insights into biological processes. These high throughput transcriptomics and proteomics measurements thus enable comparisons of clinical samples from large populations of patients and controls. This represents a formidable opportunity for the discovery of early diagnostic or prognostic biomarkers, i.e. molecular indicators of a pathological state or its probable evolution that are present before the onset of specific clinical symptoms. However, translating these fundamental discoveries into daily clinical practice is a long, tedious and costly process. In fact, in spite of intense efforts and investments, few novel biomarkers are used in clinical practice, and the rate of approved assays by the regulatory agencies is dismal. This is partly due to the lack of a linear and integrated process connecting candidate biomarker discovery to a high throughput, robust and sensitive platforms for their validation. Moreover, highly sensitive techniques are needed to overcome the tremendous complexity and the extremely broad dynamic range of protein concentrations encountered in bodily fluids proteomes. In other words, a major hurdle resides in the translation of clinical omics-technology into innovative biomarkers. Through a 4-year collaborative programme bringing together highly recognized experts in the fields of urology, clinical research, proteomics and transcriptomics, the DECanBio consortium has deployed considerable efforts in the context of bladder cancer to develop reliable quantitative analytical methods, and use them to identify new candidate biomarkers and begin to assess their predictive utility as surrogate endpoints.
To achieve these goals, the DECanBio work program was divided into four successive phases
1. Phase 1: Preparatory phase.
During this phase, the objectives were fourfold: (i) to set up the infrastructure for the collection of controlled cohort of patients with suspicion of bladder cancer, (ii) to optimize samples handling and preparation for proteomics experiments, (iii) to devise a comprehensive analytical strategy for biomarker discovery and evaluation by proteomic methods, and (iv) to start the development of a prototype of micro-ELISA system that could ultimately be used for biomarker validation.
Bladder cancer urine biobank:
The identification, selection and inclusion of the study subjects implied a very good coordination between clinical centers and the research teams to ensure both standardized conditions for sample collection as well as the inclusion of the desired number of individuals estimated in the project timeframe. At the beginning of the project, this required the development of a Study Protocol describing the details of the study design and includes questionnaires (M1.3) and standard operating procedures (M1.2) (D1.2). Its objective was to standardize methods among the participating centers and personnel involved in the study. Urine sample tracking forms were also needed to recapitulate all pertinent data required for sample and statistical analysis, including data related to sample processing and patient related data specific to the day of collection. These data were deemed necessary to provide both quality criteria for sample inclusion in the discovery step, and co-factors for statistical analysis. In addition, to satisfy legal requirements, the proper informed consent forms had to be compiled, and the protocol had to be submitted to each hospital's Ethics Committee for approval. To ensure high quality of the collected samples, it was projected that specific recommendations regarding their diet and exercise should be made to the patients prior to their appointment at the clinic. Finally, the project required the implementation of a common database dedicated to DeCanbio including a secure web-access (M1.4).
Critical issues with urine samples:
At the onset of the project, urinary proteomics being still in its initial stages, it was considered crucial to determine the optimum urine collection and sample preparation methods. Importantly, irreproducibility introduced at these early stages of the experiment would irremediably translate down the line into greater sample variability, potentially masking important signals, and eventually preventing discovery of relevant biomarkers. Thus, important efforts were anticipated to focus on a series of issues identified as critical in urinary proteomics: (1) addition of proteases inhibitors at collection stage, (2) urine storage temperature prior to centrifugation and freezing, (3) addition of preservatives to prevent bacterial growth, (4) influence of freeze-thaw cycles, (5) optimal protein extraction method and (6) minimization of technical variability.
Integrated strategy for biomarker discovery and evaluation.
Translating fundamental biomarker discoveries into clinical practice is a long, tedious and costly process. The major hurdle lies in the capacity to trim the often large lists of putative biomarkers emerging from high throughput discovery methods into more manageable sets of markers that have genuine clinical interest. This hurdle stems from the lack of a linear and integrated process connecting candidate biomarker discovery to a high throughput, robust and sensitive platforms for their evaluation. One major objective of the DECanBio project was the definition and implementation of such an integrated pipeline and its use in the context of a search for bladder cancer biomarkers.
Development of a micro-ELISA platform with electrochemical detection
Once putative biomarker lists have been pruned to the most interesting candidates, the second obstacle to clinical transfer resides in the ability to validate them in the clinical settings against large cohorts of patients. For this purpose, mass spectrometry has not yet gained general acceptance, and the more general route is to use ELISA tests. However, classical tests are time consuming and require large amounts of reagents. We thus projected to develop and thoroughly characterize a new highly automated micro-ELISA platform with electrochemical detection to facilitate the clinical validation process.
2. Phase 2: Biomarker candidates discovery phase
A wealth of information exists in the literature or on web-based resources concerning proteins involved in bladder cancer. The project plan involved curating this information and leveraging expert advice to prioritize a list of candidates from these external sources. In addition, since they have the capacity to compare samples from large populations of patients (healthy versus disease states), transcriptomics and proteomics open attractive avenues for new biomarker discovery. The initial discovery phase was designed to yield lists of gene products (RNA or proteins) displaying differential expression between normal and diseased samples based on semi-quantitative measurements.
Literature and web-based resources mining:
To leverage as much as possible of the prior knowledge about biomarkers for the system under study, we planned a thorough mining of available literature as well as web resources such as ProteinAtlas, a publicly available database with high-resolution images showing the spatial distribution of over 14,000 proteins in human tissues, cell lines, and cancer cells.
Bladder tumour transcriptomics data mining:
Genes significantly over-expressed in bladder tumors compared to normal urothelium code for proteins that are potential urine bladder cancer biomarkers. Using previously available transcriptomics data comparing normal urothelium and bladder tumour samples, the consortium was set to generate a list of gene products showing differential expression for subsequent evaluation using quantitative proteomics methods. (M2.1 D2.1). Further efforts aimed at defining alternative splice events in tumors were prompted by encouraging preliminary results that yielded a new avenue of research within the context of DECanBio.
Candidate biomarkers discovery by comparative proteomics:
Urine is readily available in large amounts using non-invasive collection procedures. Moreover, the urinary proteome contains a great variety of proteins. For bladder cancer, urine constitutes a proximal fluid, i.e. in direct contact with the diseased organ. As such, urine can be viewed as a local sink for proteins or peptides secreted, shed or leaked from the tumor. For this reason, urine represents highly relevant diagnostic analyte for bladder cancer detection using proteomics. Studying the urinary proteome was seen as a complementary approach to yield bladder cancer marker proteins that could not be predicted by transcriptome analysis of tumors. It was thus anticipated that new bladder cancer biomarkers could be derived through comparative analysis of the urinary proteome of healthy and diseased patients. In the Discovery Phase of DECanBio, protein abundance levels in patient urines had to be compared by two complementary proteomics approaches: quantitative two-dimensional gel electrophoresis (2DE) (D2.6) and Accurate Mass and Time tags proteomics (AMT) (D2.7 M2.4). Both technologies were selected to enable the unbiased discovery of bladder cancer biomarker candidates. An important aspect of these measurements was the availability of suitable urine samples. It was thus planned to provide urine samples from patients with a confirmed bladder cancer, and controls from donors, matched by age and sex, who did not present urological pathologies.
3. Phase 3: Biomarker candidates' evaluation phase
Because of the potentially high false positive rates associated with the breadth of the methods used at the candidate's discovery stage, the putative differentially abundant gene products, and their link to bladder cancer had to be carefully evaluated prior to forwarding them to the validation stage.
Method development:
To enable high throughput and reliable evaluation of the discovered candidates, Selected Reaction Monitoring (SRM) Mass Spectrometry methods had to be developed (M4.1 D4.3) to allow detection and accurate quantification of urinary proteins. In anticipation of the large number of transitions that would have to be followed, the development of acquisition scheduling was proposed. The objective was to enable the simultaneous evaluation of over 100 proteins in a single LC-SRM run.
Quantification standards:
For optimal performance of the quantitative measurements, quantification standards were required for all SRM targets included in the study. For this purpose, concatemers of tryptic peptides selected from each candidate protein sequence, called QConCAT standards, were developed and synthetized using a biotechnological approach. (M3.1 to M3.3) (D3.1 to D3.3)
Candidate qualification:
To assess the detectability and quantitative response of each candidate biomarker issued from the discovery phase, the plan called for submitting them to a first SRM screen (M4.2) that involved spiking synthetic standards into urine prior to analysis. These standards had to be designed to carry an isotopic label in order to differentiate them from endogenous biomarkers.
Candidate confirmation and verification:
Once the candidates had been qualified, the next projected step called for a confirmation of their differential abundance between the patients populations used at the proteomics discovery stage, i.e. bladder cancer patients vs. healthy controls. This had to be performed with adequate sample standardization.
Finally, the qualified candidates were submitted to verification of their predictive ability in differentiating urine samples from cancer patients and from patients with a suspicion of bladder cancer that had been cleared by pathological examination. Again, adequate sample standardization was required.
4. Phase 4: Biomarker candidate pre-clinical validation phase.
The last phase of the DECanBio project aimed at a pre-clinical validation of some of the findings about bladder cancer biomarkers using the technology developed by members of the consortium. This final goal was highly ambitious, and has unfortunately not been fully realized. As a matter of fact, delays accumulated throughout the two initial phases prevented the delivery of the list of verified biomarkers on time. In addition, benchmarking of the micro-ELISA technology yielded results that were below expectations in terms of reproducibility, which precluded its successful use in a clinical setting. The project management board, taking into account these factors, decided to entirely reorganise this final phase to allow preliminary validation of a selection of scientific results that were considered highly original and having strong potential for subsequent clinical applications. These included a set of candidate biomarkers having emerged from metallo-protein enrichment studies (D6.2 D6.3) and new potential bladder cancer detection methodology based on the study on cancer-specific alternate splice events.
Development and validation of micro-ELISA methods:
In spite of the concerns that had emerged from the technology benchmarking results, additional test were require to evaluate: (1) the analytical performance of micro-ELISA assays (including trueness, repeatability, and reproducibility), (2) their measurement range, and (3) the pre-analytical variation (including sample collection and handling, and the impact of physiological parameters (age, gender, weight, time of urine catch, etc.). For this purpose, urine samples of healthy controls and patients with bladder cancer from the discovery cohort were considered as analytes.
Pre-clinical validation of diagnostics biomarkers:
Among the proteins enriched through the metal-ion affinity fractionation strategy developed in the context of DECanBio, a set of proteins were considered of higher interest for further investigation, based on their functional annotation, and the lack of reported association with bladder cancer in earlier studies of urine.
Pre-clinical validation of prognostic biomarkers:
New findings that had emerged from the project relate to the existence of alternative splice isoforms that were found to be bladder cancer specific, and the possibility to detect these isoforms in cancer patients' urine. On the basis of transcriptomic data, the molecular taxonomy of bladder cancer could be reconsidered as suggested by Sjodahl et al. (A molecular taxonomy for urothelial carcinoma. Clin Cancer Res 2012).
Classification as Assay Specific Analytes:
It was initially planned that all biomarker assays, including MRM procedures, QconCATs and ELISAs, that would demonstrate their effectiveness during qualification, verification and/or validation steps, would be submitted for classification as Analyte Specific Reagents. Thus, provided accuracy and safety, they would be available for research use, while their potential clinical utility was being explored.
Project Results:
1. Phase 1: Preparatory phase:
a. Bladder cancer urine biobank:
The identification, selection and inclusion of the study subjects implied a very good coordination between clinical centers and research teams to ensure both standardized conditions for sample collection as well as the inclusion of the desired number of individuals estimated in the project timeframe. One hospital in France (H. Mondor in Creteil) and one in Spain (H. Mar in Barcelona) have been hired to cover: (i) patient identification selection and inclusion, (ii) data collection through both interviews and reviews of clinical charts, (iii) sample collection, and (iv) patient follow-up. The Study Protocol has been developed. This document described the details of the study design and includes the questionnaires and SOPs. Its objective was to standardize methods among the participating centers and personnel involved in the study. This study got the institutional review board (IRB) approval from H. Mondor and H. Mar. The Inform Consent forms have been designed and used in these hospitals.
Urine collection SOP: Taking into account preliminary results obtained by Partner P1 (CEA) in the context of WP2, recommendations for urine collection prepared in year 2007 through a proteomic fluid initiative coordinated by the French national institute of Cancer (INCa), and the clinical constraints; clinicians, epidemiologists, methodologists and proteomic researchers agreed on a Standard Operating Procedure. This SOP covers the volume collected, time of collection, delay elapsed between collection and freezing, storage temperature before freezing, protease inhibitors adjunction, and urine controls at bedside (hematuria, proteinuria, leucocytes, nitrites, diabetes) (M1.2). It was strictly followed by the urological centers involved in the study. Furthermore, a list of recommendation to the patients prior to collection and a questionnaire requesting information on the conditions of urine collection, have been developed. They have been translated in Spanish and French (M1.3). Also, questionnaires on epidemiological, clinical and follow-up data have been designed and translated to French from the Spanish version. A Web computerized data entry link to a database located on a CNIO server has been developed by the Bioinformatic Unit-INB at CNIO, Spain (M1.4).
Collection Plan: The collection plan for the various cohorts was finalized during the first period and periodically reevaluated. The project requirements called for three cohorts, the first for the candidate discovery studies (D1.3) the second for evaluation purposes (D1.4) and the last for the validation effort (D1.5). For the discovery phase, the definition of the population to be analyzed (90 samples) and discussions related to clinical organization (including the rate of patients actually diagnosed in the participating centers) allowed to plan the collection of the initial 90 samples that were used by Partners P1 (CEA) and P2 (BRFAA) during the discovery phase. To ensure the achievement of 60 urines samples in patient with bladder cancer (30 incident and 30 prevalent cases), it was estimated necessary to include at least 600 patients submitted to cystoscopy.
b. Critical issues with urine samples:
Urine proteome does not present a stable pattern of proteins; it varies depending on the sampling procedure (time, conditions, storage and handling) as well as the age, gender, nutrition, presence of disease, drug administration of the donor. Despite these inherent difficulties in the analysis of the urinary proteome, protein markers for diseases (including bladder cancer) have been detected in urine and have been approved to be utilized as adjuncts for disease diagnosis and prognosis. Therefore, an important task of DECanBio partners during the first period was to develop standardized approaches for urine sample handling, storage and processing (D1.2 D2.5).
In a clinical setting, urine collection may involve spiking the sample with proteases inhibitors and cold storage (0-4°C) for an extended period of time (up to 6 hours) before low speed centrifugation to eliminate cellular debris. However, considering that proteases have ample time to operate at 37°C during urine retention in the bladder, it was questionable whether adding inhibitors after collection resulted in significant changes in the urinary proteome. Furthermore, knowing that many proteins (e.g. uromodulin) do precipitate at cold temperature under salt conditions, the storage temperature before centrifugation had to be tested also. Therefore, in the context of DECanBio, a dedicated series of experiments have been design to investigate the use of protease inhibitors and the storage temperature. Several metrics were used to evaluate the influence of these pre-analytical conditions: (1) the total number of peptides identified per condition; (2) the median abundance of detected peptides per condition; and (3) the correlation coefficient between peptide abundances measured in treated samples vs. controls. All the results converged to the conclusion that the most favorable conditions for urine storage after collection were to add proteases inhibitors immediately after collection, avoid cooling the sample and, most importantly, to process them in a timely manner.
c. Integrated strategy for biomarkers discovery and evaluation.
As pointed out by Rifai et al. (Nature Biotechnology 2006, 24: 971-983), a major hurdle in putting biomarkers to clinical use lies in our incapacity to trim the extensive lists of putative biomarkers emerging from high throughput discovery studies into smaller sets of markers that have genuine clinical interest. This hurdle stems from the lack of a linear and integrated process connecting candidate biomarker discovery to a high throughput, robust and sensitive platforms for their evaluation.
d. Development of a micro-ELISA platform with electrochemical detection
One long term objective of DECanBio was to develop a microfluidic ELISA test (μELISA) system adapted to the detection of a few verified DECanBio cancer biomarkers in urine samples. Microfluidic ELISAs would offer many advantages over conventional microtiter plate assays in routine analysis in the doctor office since they will use less reagent material and provide more rapid testing. To this end, the DiagnoSwiss Company has adapted its original electrochemical microsystems by producing new batches of microchannel arrays and by coupling them with magnetic nanoparticles. One key element of this system was to enable to process samples of large volume (namely up to a few hundreds of μL) compared to the few hundreds of nL of the microchannel, and hence to pre-concentrate biomarkers on the magnetic nanoparticles, thereby increasing the sensitivity of the system. During the second reporting period of DECanBio the development of the microfluidic chip platform was pursued. This has led to the development of the ImmuSpeed, a prototype of industrial instrument (D5.4) which includes a pipetting unit, four positions for plates, two shakers, one washing station and one module for beads pre-concentration and detection in DiagnoSwiss microchip consumables. The system further features an internal personal computer and its proprietary software that enables to create assay protocols, to control the operations of the instrument and the multiplexed electrochemical detector and to process the row measurement data. The ImmuSpeed system has been dimensioned so as to have a capacity of up to 96 data points processed in series of 8 parallel analyses from standard and sample solutions embarked in a microtiterplate. The reagents are placed in the other plate positions of the machine and the microchip is regenerated between two series of experiments.
In conclusion, if the micro-ELISA approach showed good performances in terms of sensitivity, detection time and reagent volume, the developed platform was still at a prototyping stage, and the results obtained in urine samples were not robust enough to enable transfer to clinical biological practice (lack of reproducibility and difficulty in handling magnetic beads and urine samples). For these reasons, the prototype could not be used for the validation phase as initially planned, which prompted the development of an alternative plan for the final phase of the project.
2. Phase 2: Biomarker candidates discovery phase
Recent investigations have demonstrated that the urinary proteome contains a much greater variety of proteins than previously recognized. For bladder cancer, in contrast to blood, urine is a proximal fluid, i.e. directly in contact with the diseased organ. As such, urine can be viewed as a local sink for proteins or peptides secreted, shed or leaked from the tumor. In addition, urine is readily available in large amounts using non-invasive collection procedures. For all these reasons, urine represents an ideal diagnostic analyte for bladder cancer detection using proteomics strategies. The Biomarker Candidates Discovery phase of DECanBio consisted in a multi-pronged search for potential biomarkers of bladder cancer. For this purpose, literature and web resources mining, transcriptomics, and proteomics strategies were put to use to try and come up with consistent sets of proteins that should be considered for further evaluation.
a. Literature and web-based resources mining.
The DECanBio fits in a broader context of bladder cancer research, and a wealth of data had previously been accumulated on this topic in the literature using a variety of experimental approaches. The project plan called for a thorough evaluation of existing resources to select and prioritize sets of proteins previously considered characteristic of bladder cancer in the hope of detecting them in patients' urines.
b. Bladder tumour transcriptomics data mining
Genes significantly over-expressed in bladder tumor samples compared to normal urothelium code for proteins that the tumor might shed in the urine, and thus constitute potential bladder cancer biomarkers. A set of transcriptome data was available to DECanBio partners: it was obtained by Partner P4 (IC) and Partner P6 (AP-HP) in collaboration with AstraZeneca. In this series of data, the transcriptome of normal tissues and had been studied on the Affymetrix U95A array. One of DECanBio objectives was to assess these existing data using dedicated statistical methods taking into account tumour heterogeneity. This was expected to lead to the identification of roughly 200 differentially expressed genes. We first searched for the 300 most significantly overexpressed in bladder tumors (n=80) compared to normal urothelium (n=5) using the SAM software. We then examined visually all the 300 profiles. Strikingly, many of the genes overexpressed in tumors were also overexpressed in the chorion or the smooth muscle, suggesting that the overexpression observed in tumors could be not due to the transformed cells but to the stroma of the tumor. We decided to retain only genes which were significantly overexpressed in tumors compared to normal urothelium and also overexpressed in more than 25% of tumors compared to the chorion and to the smooth muscle. In addition we have also added a few genes which were obtained by specifically examining the expression profiles of family of genes which could play a role in cancer (Interleukins, CXCLs, IGFBPs, IGF). Among these genes, those which were expressed in any of the normal or inflammatory tissues were selected as new bladder cancer biomarker candidates. This approach led to a total of 80 genes whose products were considered for further evaluation (D2.1).
To complement this work, Partner P4 (IC) proposed to the Project Management Board to consider a new list of bladder cancer biomarker candidates that would correspond to splice variants specifically expressed in bladder tumors. This was approved as an extension to the initial DECanBio workplan. As a matter of fact, Partner P4 had started to analyze the expression data of bladder tumors, normal urothelium, chorion and smooth muscle on an Affymetrix exon 1.0 array. In classical chips, when they are several transcripts per gene, the measured expression is the total expression of all the isoforms as the probes are in general chosen to be common to all isoforms. In the exon arrays, as they are probes for most of the exons of the gene, one can measure individually the exons and therefore can try to infer splicing events. The idea was to find splicing events which 1) were tumor specific (not find in the normal samples), 2) would lead to the addition of new amino acids in the proteins or deletion of amino acids. Both addition and deletion would change the sequence of the proteins and therefore should lead to 1) new peptides after protease digestion which could be detected by mass spectrometry; and 2) create new epitopes which could be detected using antibodies. This work resulted in a list of 55 new peptides that should appear in the proteins as a result of these splicing events. In addition, a patent was submitted by Partners P4 (IC SR) and P6 (AP-HP) in December 2011, both on the markers obtained based on classical arrays and the markers obtained using the exon arrays (Urinary markers for monitoring bladder cancer).
c. Candidate biomarkers discovery by comparative proteomics
Looking directly at the urine proteome is a complementary approach that could be expected to lead to the identification of bladder cancer marker proteins that could not be predicted by transcriptome analysis of tumors. Therefore, new bladder cancer biomarkers were expected from the thorough analysis of the urinary proteome of healthy and pathological conditions. In the discovery phase of DECanBio, protein abundance levels have been determined by quantitative two-dimensional gel electrophoresis (2DE) and by using accurate mass and time (AMT) tag proteomics approach. Both technologies enabled the unbiased discovery of bladder cancer biomarker candidates.
i. 2DE-based discovery
The first methodology employed in DECanBio for the identification of potential bladder cancer biomarkers in urine was based on the application of comprehensive electrophoretic approaches. In brief, preparative electrophoresis in combination to 2-dimensional electrophoresis was used for the efficient and high resolution separation of urinary proteins.
Preliminary work focused on down-scaling and optimizing the prep-2DE technique so as to make it applicable for the analysis of urine samples from bladder cancer patients. Our specific aim was to optimize the various experimental steps involved in the workflow including desalting/purification of protein fractions, conditions for 2DE analysis as well as detection system, so as to avoid, the potential decrease in protein resolution imposed by down-scaling.
The subsequent step consisted in the development of a gel-based urinary proteome database. This work was performed in collaboration with the EuroKUP (European Kidney and Urine Proteomics COST Action) consortium. Specifically, two normal urine samples corresponding respectively to pooled samples from male and female healthy volunteers, were collected within EuroKUP in large quantities and distributed to multiple laboratories with the objectives to be characterized comprehensively by multiple state-of the art proteomics methods and be used as standards in clinical proteomics applications. Partner P2 (BRFAA) generated the 2DE proteomic map of this sample which included in several cases information on protein post-translational modifications. In addition a side-by side comparison of the 2DE and shotgun data collected by Partner P1 (CEA) was conducted, clearly revealing the individual advantages and complementarities of the different techniques.The results of this multi-institutional analysis have been published in the journal Proteomics Clinical Applications.
ii. AMT tag discovery
Partner P1 (CEA) has long been involved in the development and implementation of the Accurate Mass and Time (AMT) tag methodology, a high throughput quantitative proteomics strategy. The first phase of this strategy involves compiling a so-called AMT tag database of identified peptides and their parent urinary proteins from patient's urine samples having been subject to various biochemical pre-fractionations. The second phase allows the comparison of peptide abundances between large numbers of samples, as required for the comparison of patient cohorts. AMT tags proteomics thus appears very promising for candidate biomarkers discovery, insofar as it facilitates high-throughput extensive proteome analysis. This strategy was used in the context of DECanBio for label-free comparative proteomics study of urine from bladder cancer patients vs. controls.
An AMT tag database has been compiled throughout the DECanBio project (D2.4) using a variety of urine samples from diverse origins. These samples included healthy donors, patients with suspicion of bladder cancer that was subsequently cleared, and patients with non-invasive bladder cancer corresponding to various risk factors. This task required the implementation of a variety of sample fractionation techniques such as the Proteominer, and a large analytical effort. The final version of this database contained data acquired in over 1,200 LC-MS/MS analyses of urine samples. This database constitutes an in-depth repository of all peptides previously detected in urine by our group and their corresponding proteins. In addition, it contains information about peptides MS/MS spectra, which could be queried to determine lists of transitions to monitor during the biomarker evaluation phase. The final version of our AMT tag database contained over 18,000 non-redundant peptide entries, representing approximately 2,000 urinary proteins (D2.4). This AMT database was subsequently used in the DECanBio project to identify candidate biomarkers for bladder cancer, but it could also be put to use for any disease in which analysis of the urinary proteome might be relevant, e.g. to identify markers of kidney disease. Implementation of generic tools was one of the objectives of DECanBio; the AMT database of the urinary proteome can be seen as one such tool. It is publicly available on the DECanBio website (http://www.decanbio.eu).
3. Phase 3: Biomarker candidates' evaluation phase
As pointed out earlier, the breadth of the methods used at the candidate's discovery stage was associated with potentially high false positive rates associated. This was unavoidable at this stage due to the low statistical power associated with the multidimensional measurements required for the wide screening process. Therefore, the putative differentially abundant gene products, and their link to bladder cancer had to be carefully evaluated prior to forwarding them to the validation stage. One of the critical objectives of DECanBio was thus the thorough evaluation using SRM assays of three series of biomarkers candidates: the first originating from literature and web-resources mining, the second from proteomics measurements, and the third from recently obtained exon-chip transcriptomic assays. This evaluation was setup in three consecutive steps: (1) candidate qualification, consisting in checking the detectability and the linearity of the quantitative response of the corresponding peptides in urine; (2) candidate confirmation, i.e. assessment of the differential concentration of a candidate between urine from healthy donors and urine from cancer patients, thereby confirming, when available, the results obtained during the discovery phase; (3) candidate verification, involving the accurate estimation of the differential concentration of the candidates in urines from cancer patients vs. urines from clinical controls having a suspicion of bladder cancer and for whom the suspicion had been cleared by pathological examination. Before this work could be undertaken, significant efforts had to be devoted to the development and optimization of methods and standards to achieve the evaluation task.
a. Method developments.
A proteomics screening experiment is performed at peptides level, and the selection of targets uniquely associated to the protein of interest. In contrast to conventional shotgun proteomic studies, selected reaction monitoring (SRM) measurements are quantitative, rigorously targeting a predetermined set of peptides. The determination of the proper SRM transitions for each target is an essential step in biomarker evaluation. More precisely, for each protein of interest, the peptides presenting good ionization properties and which are unique surrogate of the protein (or a specific isoform), called proteotypic peptides (PTPs), need to be selected. For each peptide, fragment ions that exhibit strong signals in MS/MS spectra and discriminate the targeted peptide from other species were preferred. In addition, other parameters critical to SRM analysis, such as collision energy also required optimization.
b. Quantification standards
As initially evocated in the technical risks assessment of DECanBio DoW, the protein content of urine samples varies depending on the age, sex, diet, medication, time of sample collection etc. All these factors may complicate the interpretation of results when dealing with biomarker discovery. One way to take these factors into consideration is the adjustment of the protein concentration of the urine samples under comparison. This was achieved by protein concentration measurement using pyrogallol red and Bradford assays. In addition, for absolute quantification of urinary proteins, the adjunction of quantification standards was a necessary requirement. We opted for QconCAT standards, concatemers of tryptic peptides selected from each candidate protein sequence.
The project thus required the synthesis of multiple QconCAT polypeptides covering a list of potential biomarkers for the diagnosis of incident and recurrent bladder cancer against a background of healthy subjects and patients with other, non-cancerous bladder-related complications. Each QconCAT consists of multiple isotope-labeled proteotypic peptides. Each QconCAT could be spiked into a biological sample, and after digestion with trypsin, the contained reference peptides were released, as well as their counterpart peptides from the natural proteins present in the sample. By comparing peak intensities in a mass spectrum of natural and reference peptides, the original amount of native protein present in the sample could be determined for all proteins covered by a QconCAT.
Partner P7 (Entelechon) was in charge of the development and production of isotopically labelled standard polypeptides (QconCAT's) to be used by Partner P10 (CRP-Sante) for the detection and accurate quantification of bladder cancer biomarker candidates in urine samples by mass spectrometry. This included the bioinformatic optimisation of the expression constructs for the concatamerisation of the selected putative proteotypic peptides, the gene synthesis of these constructs, as well as the cloning, expression and labeling of the QconCAT standards.
c. Candidates qualification
A critical part of biomarker candidate evaluation consisted in the development and the implementation of a workflow to translate a protein/transcript list of candidates into peptides used to perform a wide screen using the MS-based analysis in urine. It had a first objective to detect and identify the peptides (and per association the proteins) in urine samples. This was called the biomarker candidate qualification phase.
The objective of this work was to confirm the detectability in urine samples of the proteins from three lists of prioritized bladder cancer biomarker candidates: The first priority list, called Set 1 containing 152 protein candidates, had been selected based on their biological and clinical relevance. The selection of these candidates was a compilation of results from the scientific literature, combined with protein, for which a bladder cancer specific expression profile was observed based on immunohistochemistry patterns for a large number of human tissues, cancers and cell lines, available through Human Protein Atlas (www.proteinatlas.org). In addition, 80 proteins corresponding to genes, whose expression was shown to be dysregulated in tumors by Partner P4 (IC-SR) were appended to this list. A second set of biomarker candidates, called Set 2 encompassing 55 proteins, was composed of proteins emerging from two biomarker discovery studies (Task 2.3 and 2.4) performed by Partners P1 (CEA) and P2 (BRFAA). Finally, a third set, Set 3, was obtained from transcriptomic data, and with a focus on skipped/added exon data by Partner P4 (IC SR). Select proteins from these three sets of candidates were subsequently evaluated using robust instrument methods for SRM analysis developed previously.
Protein Candidates Set 1: A series of 33 proteins (Top 33) resulting from the confrontation of Set 1 with the urinary proteome AMT database (D2.4) created by partner P1 (CEA) within, formed the basis for a primary SRM screen. Eighty six peptides, surrogates of the 33 protein candidates, were unambiguously detected in patient urines and their identity confirmed using stable isotope labeled synthetic peptides. A second set of 41 proteins from the 152 entries priority list of candidates had been selected and screened. Sixty eight peptides, surrogates of 37 protein candidates, were unambiguously detected and confirmed with the stable isotope labeled peptides, while four proteins were not detected by monitoring the corresponding peptides. Three additional proteins from Set1 and eight other new proteins candidates were selected based on their clinical relevance and new evidence gathered from studies that became available within the project timeframe (literature mining, transcriptomics data). In total, 82 out of 152 protein candidates from Set 1 were qualified during this stage of the study (D4.4). The results of this screening experiment led to the design and synthesis of four isotopically labeled QconCAT standards by Partner P7 (Entelechon).
Protein Candidates Set 2: The screening of 55 protein candidates from the proteomics Discovery phase of DECanBio performed by Partner P1 (CEA) and P2 (BRFAA) in the context of the discovery phase was performed. From these 55 candidates, 17 were already screened in Set 1. One hundred seven peptides, surrogates of the remaining 38 protein candidates, were unambiguously detected and confirmed using stable isotope labeled peptides. The results of these experiments have led to the design of two additional isotopically labeled QconCATs standards that were synthetized by Partner P7 (Entelechon).
Protein Candidates Set 3: The candidates in Set 3 were constituted by 19 gene chip exons related to 21 gene candidates. The amino acid sequences (protein level) resulting from the translation of the gene exon candidates (mRNA), termed proteo-exon were selected. For each proteo-exon, a tryptic sequence covering the variable region (added or skipped sequence string) was selected to be monitored, together with five unique canonical peptides per protein. From the 19 original gene chip exons corresponding to 21 gene candidates, 9 proteo-exons related to 8 gene candidates were selected in accordance with SRM detectability criteria (e.g. amino acid composition, hydrophobicity, sequence length, etc). Upon data evaluation, two cases were observed: i) the detection of the proteo-exon and the canonical proteotypic peptides from the same protein candidate and ii) the sole detection of the proteo-exon, without the corresponding canonical sequences, suggesting the existence of other modifications on the corresponding gene products. The outcome of this screening resulted in the design of a QconCAT standard that was commissioned to Entelechon (FP7 Partner).
d. Candidate confirmation and verification
The subsequent phase of the work was to perform the quantification of the 134 qualified biomarker candidates in clinical urine samples from the so-called verification cohort (D1.4) in order to evaluate the differential abundance between physiological and pathological concentration, using high throughput SRM methodologies that were setup previously.
One hundred twenty three samples provided by partners P6 (AP-HP) and P9 (CNIO) were received and processed (peptide level) using the standardized protocol established by partners P1 (CEA) and P3 (ETH). This collection included urine samples from few healthy individuals, from patients with urological disorders for whom the cancer diagnosis had been cleared (real-life controls), and from bladder cancer patients stratified according to their EUA risk group (low, intermediate, and high), with associated demographic and clinical information. Four categories of patients could thus be distinguished: Incident: Patients with a first confirmed incidence of Bladder cancer; Incident control: Patients with no history of bladder cancer diagnosed with other urological diseases, Prevalent: Patient with a prior history of bladder cancer who, during their clinical examination, were diagnosed with bladder cancer recurrence; Prevalent control: Patient with a prior history of bladder cancer, whose examination results cleared any suspicion of bladder cancer recurrence.
4. Phase 4: Biomarker candidate pre-clinical validation phase.
The pre-clinical validation phase was originally planned as a way to transfer the list of urine bladder cancer biomarkers emerging from the discovery and evaluation phases to the clinic through immunoassays validated on independent cohorts, with clinical data and urine samples prospectively collected in the context of the project. During the DeCanBio project, considering recent development of SRM technologies, and the difficulties in obtaining immunoassays of good reproducibility in urine, the management committee of DeCanBio decided to perform the pre-clinical study of candidate biomarker through SRM. In addition, it appeared later in the project that the use of specific transcript splicing events could be a new source of biomarkers. As recently analysis of RNA extracted from urine has been shown to be feasible on a large scale, the consortium therefore started to develop assays to directly measure these splice forms in the urine.
During the project s third period, it became acutely evident that the rescheduling of series of tasks (onset of collection delayed due to changes in Spanish legislation, subsequent postponing of cohorts deliverables, rescheduling of the candidate discovery phase) that had accumulated since the beginning of the project could not be expected to be recovered during this final period. . In addition, the list of candidate biomarkers being still at the evaluation stage, it was not possible yet to proceed to their pre-clinical validation. The project management board thus decided to explore alternative ways to validate other findings obtained in the context of the project instead of the initially planned pre-clinical validation of the verified candidates. As a matter of fact, several unanticipated results, that could have significant impact on the understanding and/or diagnosis of bladder cancer, required additional proof.
a. Preliminary validation of candidates emerging from the metallo-protein screen.
Among the proteins enriched through the IMAC fractionation strategy developed by Partner P2 (BRFAA) during the preparation phase of DECanBio, aminopeptidase N, myeloblastin and profilin 1 were considered of higher interest and therefore prioritized for further investigation, based on their functional annotation, and the lack of reported association with bladder cancer in earlier studies of urine. In addition, during period, IMAC fractionation was combined with capillary electrophoresis- mass spectrometry (CE-MS; in collaboration with Dr H Mischak) for the study of the low molecular weight proteome (peptidome). This study was an expansion of an existing collaboration applying CE-MS to identify urinary peptides associated with bladder cancer staging. This study identified four urinary peptides, derived from uromodulin, collagen type I and III, and membrane-associated progesterone receptor, that enabled, in combination with cancer grade, the separation of muscle-invasive bladder cancer from other cancers, with had shown a negative predictive value of 77% and a positive predictive value of 90% in a previous blinded evaluation (Schiffer and Vlahou et al., Clin Cancer Res 2009).
b. Preliminary validation of splice variant results
The project management board decided to redirect validation efforts toward the assessment of some highly innovative findings made during the final period. These findings relate to the existence of alternative splice isoforms that were found to be bladder cancer specific, and the possibility to detect these isoforms in cancer patients' urine.
As a first step, Partner 6 (AP-HP) performed systematic immuno-histochemical studies using antibodies targeting some candidates potentially associated with the molecular taxonomy, among them cytokeratin 18, cytokeratin 20, Topo2a, PI3, FoxA1, S100A7, S100A8. Protein expression was assessed using tissue microarrays gathering the cases previously studied at transcriptomic level. For most of the candidates, the correlation between mRNA and protein expression was good or excellent. Preliminary analysis according to the stage and the grade showed frequent association with the expression. Further investigation will aim to propose a combination of biomarkers covering all the sub-phenotypes observed in bladder cancer, in order to test them in urine and estimate their positive and negative predictive values.
Potential Impact:
According to current clinical practice, the standard care for the detection of bladder cancer is cystoscopy and pathologic examination. Thus, urine cytology remains, despite its limitations, the standard non-invasive method for detecting bladder cancer, keeping a superior specificity to most available tumor markers. For this reason, no tumor biomarker could be recommended for use in bladder cancer screening at the onset of the project. Thus, the development of biomarkers for bladder cancer was still a matter of research. DECanBio implemented a novel workflow and unique state-of-the-art technologies: (i) to discover biomarker candidates by integrating various omics technologies; (ii) to qualify and verify these candidates using SRM-LC-MS technique which allowed to rapidly screen hundreds of candidates and to establish a reduced panel of biomarkers that could be the starting point of future validation efforts using dedicated SRM or ELISA assays. The most practical use of non-invasive tests could be directed to reduce the number of surveillance cystoscopies performed during the monitoring of patients treated for bladder tumors, and to detect precociously bladder tumor progression and dissemination by setting clinical protocols with frequent and convenient urine tests. The results of DECanBio research could indirectly affect not only the population suffering from bladder cancer but could have a major impact on the diagnostics/prognostics of multiple diseases such as the very prevalent kidney diseases, prostate cancer and others.
The scientific impact of DECanBio in terms of translational research has been very prominent as evidenced by the feedback received during the numerous presentations of the consortium's efforts in this area. In particular, the emergence of an integrated method for biomarkers discovery/evaluation based on the AMT tag methodology for the discovery phase and the iSRM strategy for the evaluation phase can be considered highly significant.
One of the critical objectives of DECanBio was the thorough discovery of urinary proteins with a potential link to bladder cancer. The corresponding discovery phase culminated in the delivery of three series of biomarkers candidates: the first originating from literature and web-resources mining, the second from proteomics measurements, and the third from recently obtained exon-chip transcriptomic assays. While a significant portion of these candidates has been subject to evaluation, we necessarily had to make a prioritization and not all candidates could be considered within the context of the project. These lists contain relevant targets for future evaluation efforts, and the consortium will have to decide how to make them available to the public for future studies.
An important task of DECanBio partners during was to develop standardized approaches for urine sample handling, storage and processing. Thus, important efforts focused on a series of issues identified as critical in urinary proteomics: (1) Addition of proteases inhibitors at collection stage (2) Urine storage temperature prior to centrifugation and freezing. (3) Addition of preservatives to prevent bacterial growth. (4) Influence of freeze-thaw cycles (5) Optimal protein extraction method (6) Minimization of technical variability. Standards operating procedures have been developed and published in the scientific literature, helping set guidelines for future urinary proteomes studies, not only in the context of bladder cancer, but for a variety of other pathologies.
A highly significant outcome of the DECanBio was obtained at the candidate evaluation phase, yielding a panel of 24 urinary proteins distinguishing bladder cancer patients from urologic controls. This panel constitutes the starting point for a future bladder cancer test, providing adequate validation efforts are undertaken. Mandate will be granted to Partner P7 (Entelechon), in consultation with the project coordinator and other members who have expressed their interest, to approach a diagnostic company interested in building a clinical diagnostics assay on the obtained results.
Another very important result that emerged from DECanBio project relates to the new concept for cancer detection involving spliced exons. The measurement of splicing events instead of global expression level of genes could significantly improve the specificity and the sensitivity of bladder cancer detection either for the initial diagnosis (patients presenting with hematuria) or for the follow-up. Furthermore, upon the development of ELISA or SRM assays specific of the spliced sequences, the splicing variants will also become measurable at the protein level as evidenced by their detection during the biomarkers evaluation phase of DECanBio. A patent was filed by Partners P4 (IC SR) and P6 (AP-HP) on this novel concept.
An important exploitable foreground generated in the frame of the DeCanBio project refers to the work of Partner P8 (Diagnoswiss) on the newly developed electrochemical microfluidic chips, the prototype of robotic analysis platform and specific test protocols for bead-based immunoassay applications. This foreground aims at obtaining a finalized first-series instrument applicable for beta-testing in view of finding a strategic partnership for the commercialization of a novel automated Elisa instrument or the transfer of the technology for R&D applications. The development of microfluidic systems as well as novel approaches reducing the number of assay steps (and hence of manipulations) opens new routes for the application of immunoassays in research laboratories and medical diagnostics, with the benefits of reducing the analysis time and cost. The results of the DeCanBio project confirm that the microfluidic chips, the robotized infrastructure and the developed assay protocols enable to perform unattended bead-based immunoassays with electrochemical detection that can be applied to multiple sequential analysis of specific proteins or biomarkers. In addition to the automation of the assays, the main advantages of the developed micro-ELISA technology rely on the reduction of the reagent consumption (and hence of the unitary test cost) and on the shortening of the time-to-result compared to traditional microtiter-plate methods. Application to urine samples has yet shown limitations due to the complexity of the sample matrix which interferes with the formation of the immune complex or induces strong background masking the effectively measurable signals. If the assay development work shows the flexibility of the developed electrochemical microchip platform in terms of applicable assay types and analyte concentration range, more relevant biochemical reagents are required to improve the selectivity and reproducibility of the approach with urine samples. Otherwise, further expertise will still be required to improve the handling of the magnetic beads and of the very low volumes involved at each assay step, and to better assess the robustness and long-term use of the platform in real conditions. Such additional work is envisaged for transforming the foreground generated in this project into a final product adapted to routine analysis in R&D or medical diagnostic laboratories.
In terms of direct social impact, DECanBio has resulted in the hiring of ten personnel (five women and five man) in positions entirely dedicated to the project. They included among others:
- A technician biochemist (Ms Madalen LeGorrec) and a computer scientist (Ms Claire Adam) by Partner P1 (CEA) to perform sample preparation and AMT tag database compilation and management respectively.
- An engineer statistician (Mr Mourad Mellal) and a post-doctoral associate (Dr Aurelie Kamoun) by Partner P4 (IC SR) to perform statistical analyses of AMT tag data and splice variant data respectively.
- A post-doctoral associate (Dr Elodie Duriez) for SRM analyses by Partner P10 (CRP-Sante).
- An assistant engineer (Mrs Edwige Lopez-Perreira) to supervise urine samples collection at Partner P6 (AP-HP).
In terms of advancing higher education, the project has been central to the PhD thesis of Mrs Magali Court and the Diploma work of Mr Mourad Mellal at Partner P1 institution (CEA). The PhD thesis topic aimed at the characterization of the urinary proteome, and of its potential for disease biomarkers research. The diploma work focused on the use of penalized logistic regression for the interpretation of comparative proteomics data.
In conclusion, the work performed within DECanBio led to several highly important outcomes. Among these, the determination of biomarker candidates for bladder cancer diagnosis and recurrence/progression constitutes a remarkable achievement. Provided it is followed up by appropriate validation efforts, these results could have major economic and sociological impact through: (i) decrease the number of the costly and highly uncomfortable and unnecessary cystoscopies; (ii) schedule in a more efficient way the surgical intervention (cystectomy) which severely compromises the quality of life of patients; (iii) guide the design of effective therapeutic interventions.
List of Websites:
http://www.decanbio.eu