Final Report Summary - DIPROMON (Multimodular biomarker analysis workflow for diagnosis, prognosis and monitoring of drug treatment response in bladder cancer)
Executive Summary:
Background:
Bladder cancer (BC) is a prevalent malignancy imposing a major burden on patients and healthcare. While first line procedures are successful in reducing tumor load, BC is characterized by a high probability for recurrence, eventually developing into muscle invasive disease. In consequence tight clinical monitoring is implemented with cystoscopy as gold standard.
Challenge:
As with most clinical diagnosis procedures sensitivity as well as specificity see limits also in BC, hence improving BC management via adding supportive, ideally non-invasive procedures sees clear clinical need. DIPROMON was set out to improve diagnostic accuracy for recurrence and disease progression via identifying a panel of molecular biomarkers. Yet the biggest challenge is identification of appropriate measures triggered by such diagnosis – i.e. biomarker guided precision in drug use.
Integrative concept:
For adding to precision medicine exemplified on BC we followed a concept of modelling disease pathology and drug mechanism of action as molecular process models, and selected biomarker candidates at model interference. Including experimental testing on disease progression-associated molecular processes promises identification of biomarker candidates with on top of diagnosis/prognosis also support in drug selection as downstream consequence.
Technology implementation:
Along DIPROMON procedures we first established BC pathology models including annotation regarding association with disease progression, from there picking a panel of biomarker candidates. Assembling experimental materials further triggered assay development including multiplexed ELISAs and cell-based analytics (morphology and surface marker stains), embedded in a software framework for computing classification functions on risk for recurrence/disease progression.
Results:
On the basis of an established DIPROMON biobank we evaluated biomarker-based classification functions for BC recurrence diagnosis, a top-ranked classifier forwarded to an independent validation study. AUC values obtained indicate capturing a relevant portion of BC recurrence events. Involved biomarkers – due to their assignment to defined molecular processes – allow analyzing interference to given therapy regimes (as BCG), and further allowed us screening alternative drugs promising beneficial interference with BC pathology.
IP and dissemination:
All relevant findings of DIPROMON were published in peer-reviewed journals and presented at scientific conferences for assuring provision of our results to the scientific domain. Specific aspects were filed for IP protection.
Outlook:
A blueprint of the integrated DIPROMON analysis platform including SOPs for all procedures is in place and ready for further validation studies.
Project Context and Objectives:
1.2 Description of project context and objectives
1.2.1 Bladder cancer diagnosis and therapy – clinical perspective
Urinary bladder cancer (BC) is a common disease worldwide. About 2.7 million people have a history of BC. The highest incidence of BC is in developed countries, especially in North America and Europe. In USA alone, about 570,000 people suffered from bladder cancer in 2011, with steadily increasing prevalence. The WHO estimates that there are 330,000 new cases annually worldwide. In Germany, about 26,000 new cases of bladder cancer are diagnosed every year.
The most common clinical presentation is blood in the urine/hematuria, which is the principle initial symptom in 90% of bladder cancer patients. Usually this is painless and the blood may be visible to the naked eye (gross hematuria) or can be seen only under the microscope (microscopic hematuria). Frequently, the diagnosis of bladder cancer is delayed because bleeding is intermittent or attributed to other causes, such as urinary tract infection or blood thinners. However, a substantial proportion of these patients will have a significant problem such as kidney stones or tumors, urinary tract obstruction and bladder cancer.
Bladder cancer has high survival rates but also high recurrence rates, which leads to a requirement for an active surveillance regime for people with a history of bladder cancer. Even after full excision of superficial bladder tumors, there is up to a 75% chance of recurrence. Although most recurrences are also superficial, bladder cancer is a multi-focal disease and new tumors may not necessarily occur at the same location. Accordingly, a post-surgery monitoring program is always initiated. Guidelines from the American Urology Association recommend follow up is undertaken quarterly, reducing to annual checks at five years from first diagnosis. In most cases surveillance tests are performed by cystoscopy, a routine but unpleasant procedure for the patient.
Because of long-term survival and the need for lifelong routine monitoring and treatment, the cost per patient of bladder cancer from diagnosis to death is ranging from $96,000-187,000 in the USA. Overall, bladder cancer is the fifth most expensive cancer in terms of total medical care expenditures, accounting for almost 3.7 billion $ in direct costs just in the USA.
Current laboratory diagnosis of bladder cancer is done by examination of the urine, vagina, or rectum. The following tests and procedures are mostly used:
• Urine analysis: A test to check urine content including sugars, proteins, red/white blood cells
• Urine cytology: Microscopic examination of urine to check for abnormal cells
• Cystoscopy: A procedure to look inside the bladder and urethra to check for abnormal areas
Despite its low sensitivity (35% to 40%) in the detection of urothelial carcinoma of all grades and stages (17% in grade 1, 61% in grade 2 and 90% in grade 3), urine cytology remains the most commonly used non-invasive test. If the urinary cytology is positive, then transitional cell cancer of the urothelium is almost certainly present. However, cytologic examinations may be negative in up to half of patients with bladder cancer; thus, a negative study does not rule out bladder cancer.
In this context the objectives of DIPROMON were to add options aside invasive cystoscopy in surveillance by non-invasive cytology in combination with biomarker tests. The set of biomarkers is aimed to give information on the probability of recurrence and invasiveness of the tumor and prognosis on the effectiveness of therapy. Especially the individual response to BCG immunotherapy in respect to chemotherapy should be assessed.
1.2.2 Project objectives
With project filing we defined the following specific objectives to be tackled with DIPROMON:
• Design of a novel patient stratification concept
• Building tools and procedures for allowing implementing such stratification concept
• Validating the concept in the clinical context of bladder cancer
Patient stratification concept
The general topic being followed in DIPROMON experienced significant momentum in the translational research community during project runtime, coining “stratified” as well as “precision” medicine. The general notion is to improve molecular phenotyping on top of clinical phenotyping, i.e. adding or ideally replacing clinical descriptors of disease presentation with molecular mechanistic parameters. Reasoning is apparent: a clinical presentation exhibits a molecular pathophysiology, itself being a composite of molecular processes. A molecular biomarker in its definition serves as proxy for the status of molecular processes, hence assembling a panel of biomarkers reflecting pathologic processes of a clinical presentation serve for detailed molecular phenotyping. Such strategy holds a major benefit in the context of analyzing drug treatment and response, being the pivotal aim for improving diagnosis and prognosis (making only sense if triggering a clinical consequence). In essence, a drug modulates a target’s activity, triggering changes in downstream molecular processes. Hence both, disease pathology as well as drug mechanism of action can be described on the level of molecular processes, with molecular biomarkers as anchors. On this basis a development strategy towards realizing stratified/personalized medicine holds the following elements:
• Biomarker-based molecular phenotyping
• Clinical phenotyping (holding established factors for describing disease presentation and progression)
• Molecular models of drug mechanism of action
Combining these elements (with improved diagnosis and prognosis as inherent element) promises improved precision in drug use and molecular mechanistic guidance in drug development.
DIPROMON tools and procedures
In-silico models of pathology and drug MoA
Rich molecular data space is available for characterizing bladder cancer pathology, however, needing consolidation on the level of molecular processes for allowing rational selection of a corresponding biomarker panel. This needs an algorithmic procedure for traversing molecular feature sets into molecular model representations. Same applied on feature sets characterizing drug mechanism of action allows molecular process interference analysis for linking drug effect and disease pathology with molecular biomarkers.
Biomarker materials
Candidate biomarkers and respective antibodies need production and purification for implementing assays.
Multiplexed assays
With focus on urine as sample matrix assays have to be established, according to the biomarker panel concept needing further integration in multiplexed formats. For prototyping and subsequent clinical use automated readout technology needs to be in place.
Cell-based analytics
Both, morphological characteristics (to be analyzed via specific staining) as well as molecular characteristics (expression of relevant receptors) are to complement biomarker-based profiling of samples.
Integration software
Software is to be established taking a clinical/molecular data set as input for computing a score according to a classification function.
SOPs
For assuring standardized application of components and workflow set of SOPs is to be established
Validation, bladder cancer
For validating the concept a biobank of bladder cancer samples was to be established, including detailed clinical phenotyping at baseline (first diagnosis), medication regimes and follow-up data on results from surveillance according to clinical practise.
Establishing and validating tools and procedures was conducted in a two-step fashion, a discovery study for defining relevant clinical and molecular parameters and respective classification functions, followed by a validation study.
Primary focus in-line with clinical need was recurrence diagnosis, for reflecting our concept combined with treatment regimes further coupled with prognosis towards development of muscle invasive disease.
Project Results:
Description of main S&T results/foreground
1.3.1 Project workflow and specific goals
With focus on bladder cancer DIPROMON was organized in six RTD work packages complemented by management as well as dissemination/exploitation.
Central procedure implemented biomarker candidate selection, on such basis establishing experimental tools for marker quantitation, entering a discovery study for biomarker shortlisting and classifier generation, further forwarded to a validation study.
In parallel the conceptual fundament of our stratification approach was elaborated and continuous biomarker profiling for optimizing diagnosis/prognosis/prediction was realized.
The DIPROMON biobank provided sample grounds for biomarker and classifier testing, altogether meeting in technology prototype development and integration blueprint together with respective standard operating procedures.
With this structure we implemented the central aims of DIPROMON (design of a novel patient stratification concept; building tools and procedures allowing implementation; validation) in project runtime.
1.3.2 Stratification concept
With precision medicine and personalization considered as key strategy for improving efficacy of approved drugs as well as improving the development path in drug development, patient phenotyping on a combined clinical and molecular level in regard to the phenotype in focus is proposed, with a critical role of early stage clinical testing.
Disease pathology and drug effect resemble molecular process perturbance and respective interference with drug mechanism of action, demanding description on the level of molecular phenotyping. In preclinical models and early stage trials in human a link of molecular processes, clinical phenotype parameters and outcome via molecular biomarkers bi is required. Specifically early stage clinical trials serve for evaluation of a drug’s molecular mechanistic impact in human via including molecular process biomarkers already in study design. On such basis adaptive trial designs as biomarker-based enrichment can be pursued from phase II trials onwards, linking molecular processes to drug effect via use of predictive biomarkers.
Here, diagnosis and prognosis are to be considered intrinsically different to prediction of drug response. While expanded phenotyping may provide improved accuracy in detecting onset and progression of disease, prediction claims at adding precision in drug selection and response with respect to beneficial outcome on a personalized level. Clinical together with molecular phenotyping, in particular via utilizing molecular biomarkers, serve as most relevant tools in this context. However, diagnosis as well as prognosis is not sufficient in case we aim at controlling development of an endpoint of a molecular system S, i.e. some P(S), in the given scenario implying interference with S via utilizing a specific drug and its mechanism of action. For assuring control we need specific knowledge about the structure of S, otherwise we lack means for rational interference. Prediction rests on diagnosis of an aggregate property of a molecular system S at a certain time t (clinical follow-up) and integrates prognosis with structure (in a molecular process/pathway context). Purpose of prediction is to prognose the consequence of interference with S via a selected drug.
Consider a system S reflecting a dynamical hierarchy. Implicit question is what molecular characteristics xi to modify – including the consequences this may have on the context of xi, being the entire dynamical hierarchy, for allowing such optimization. The drug in the first place introduces an object constraint on the specific target(s) it is binding, in further consequence propagating this constraint in the dynamical hierarchy. Due to the relation constraints, consequence of addressing the target may become apparent on various levels of the hierarchy, eventually also including a change in a clinical parameter. In case the drug is applied on the level of S – although seeing molecular phenotypes Si, Sj (an appreciated fact specifically in oncology) – a further consequence may be variance in drug response, eventually detected on the level of a clinical parameter, but per design detected on the level of molecular biomarkers out of X, resembling the requirements of precision medicine.
Let us consider some S holding a drug target and two molecular biomarkers. Addressing the target via a drug (the drug imposing an object constraint on the first order object) may trigger a change in some higher level observable O2 (e.g. a molecular pathway status) further reflecting in some clinical phenotype parameter O3. In a different composition of S no consequence is seen on the level of O2, but still exhibiting an equivalent readout in O3. Such scenario reflects two molecular phenotypes Si, Sj out of S, in case S is defined on the level of O3. If defined on the level of O2 two strata of S can be determined, namely Si and Sj.
O2 in this example setting resembles the status of a molecular pathway, with biomarkers xi and xj characterizing two distinct molecular processes being constituents of the pathway. According to the given scheme the abundance (e.g. with respect to a certain threshold) of both biomarkers are indicative for O2 in the one situation, whereas just one molecular process is identified as active in the other situation (e.g. with a down-regulated xj), in consequence seeing difference in O2. Although O3 is equivalent in both cases, altering O2 may be the essential parameter regarding a clinical endpoint P(S). A practical molecular example is the apoptosis pathway with promoting (determined by xi) and inhibiting (determined by xj) molecular process components.
Aiming at interfering with a drug addressing xt, hence altering the process seen as relevant via the proxy xi, will in the example case see consequence on P(S) in Si, but no effect for subjects embedded in Sj. In consequence, we expect variance in drug response on the level of S, a fact which can be determined when including xi and xj in stratification for response when targeting xt with a certain drug.
According to this concept any S needs to be captured on the level of molecular processes. Equivalently, drug mechanism of action needs to be assessed on a molecular process level for allowing a fit-for-purpose evaluation utilizing molecular biomarkers as process proxy.
For leveraging on this concept bladder cancer together with drug mechanism of action need to be modelled on a molecular process level.
1.3.3 Bladder cancer pathology and biomarker candidates
In line with the stratification concept we consolidated a repository of molecular profiles, signatures and individual features characterizing BC, split into superficial as well as muscle invasive disease. We in total extracted molecular features from 5 SNP studies, 3 mRNA transcriptomics studies, 1 miRNA transcriptomics study, 2 proteomics studies, and 3 metabolomics studies resulting in a set of 897 relevant proteins after mapping miRNA to their targets and metabolites to enzymes. Additionally we executed literature mining (NCBI PubMed) for retrieving individual molecular features reported as relevant in BC, finally concluding on a set of about 1,300 molecular features linked with BC.
For traversing the molecular feature set into a molecular process model we mapped the molecular feature set on a hybrid protein interaction network, followed by using topological criteria for extracting molecular processes and their biological relations. Molecular features remaining in the molecular model were screened for annotation in the biomarker context (split into diagnostic/prognostic), combined with Omics profile analysis on disease progression according to statistical testing. With such biomarker annotation in hand we analyzed the molecular pathway content of each process for defining molecular processes of specific relevance in disease progression, followed by selection of representative biomarker candidates for each molecular process deemed being involved in progressive disease.
According to this procedure we concluded on a first set of biomarker candidates, complemented by a consortium efforts in identifying further candidates deemed relevant but not covered in the process model.
Further analytics focused on cell surface markers including proteins of the human epidermal growth factor receptor (HER) family, Uroplakins and CK20/CD45. The last pair is used to discriminate cancer cells from white blood cells. On top, CK20 appears also to be a prognostic marker on bladder cancer cells.
We from this analytics steps concluded on a set of about 30 biomarker candidates to be further followed materials development for ELISA and cell staining.
For complementing the deductive analysis on the molecular model basis also an explorative strategy was followed via transcriptomics profiling of samples (Affymetrix® Human Gene 1.1 ST Array Strip using Affymetrix GeneAltas System).
1.3.4 Materials and technologies
Biobanking
According to clinical needs the following key areas of interest were defined:
• Adding precision/changing procedures for BC recurrence diagnosis
• Estimating risk for disease progression, muscle invasive disease
• Estimating response to medications
Primary focus of assay/product development was on diagnosing recurrence for complementing regular surveillance procedures. Biobanking of samples combined with deep annotation of clinical parameters primarily focused on this scenario, further complemented by sampling advanced stages. Compliance with ethics requirements were explicitly handled and reported in the context of project deliverables. The full DIPROMON biobank finally included in total 249 transitional cell bladder cancer and 80 muscle invasive bladder patient samples We have also collected urine and blood samples from 114 healthy age matched volunteers as controls.
new diagnosis Recurrence Neg Neg previous pos HVC G1 pTa G1 pT1 G2 pTa G2 pT1 G3 pTa G3 pT1 TBC
urine 66 79 43 61 114 6 2 90 24 6 26 15
Blood Paxgene 7 15 1 22 112 2 0 13 1 0 1 0
Plasma 7 15 1 22 112 2 0 13 1 0 1 0
Table of samples of urine and blood from Non-muscle-invasive bladder cancer patients or control (grade and stage at sample).
G2 pT2 G3 pT2a G3 pT2b G3 pT3a G3 pT3b G3 T4a G3 T4b
6 36 0 18 1 17 2
Table of samples from invasive bladder cancer patients (grade and stage at sample)
Materials
For the spectrum of biomarker candidates a pragmatic procedure was implemented with respect to material availability for designing assays, in part including commercially available reagents. Still, significant work was involved in generating new materials including recombinant production of proteins as well as corresponding antibodies. By this a set of reagents became available being of general interest in BC research.
EN2 Engrailed 2 UPK3A Uroplakin-3a FGFR3 Fibroblast growth factor receptor 3 CD44 molecule (Indian blood group)
IL8 Interleukin 8 MMP9 Matrix metallopeptidase 9 VEGFA Vascular endothelial growth factor A
VIM Vimentin
EGFR Epidermal growth factor receptor APOE Apolipoprotein E Cyfra 21-1Cytokeratin Fragment-19 MYC Myelocytomatosis viral oncogene homolog (MYC)
PTGS2 Prostaglandin-endoperoxide synthase 2 EN2_full Length with 2 his-tags Engrailed 2 FGFR3-First N-ter domain Truncated FGFR3 FGFR3-Second C-ter domain Truncated FGFR3
IL-6 Interleukin 6 UPK1B Uroplakin 1B KRT20 keratin 20 CDH1 E-cadherin
Multiplexed ELISA
Multiplexed sandwich immunoassays on the ARChip Epoxy platform were established to quantitatively measure protein biomarkers. Analyte binding was detected with biotinylated antibodies in combination with Streptavidin/Dy647. Assay conditions were optimized with respect to the chip platform (ARChip Epoxy), printing buffer (PBS at pH 7.2 supplemented with Na-deoxycholate), probe concentration (capture antibodies spotted on the chip), assay buffer (commercial available LowCross buffer) and signal output.
Recovery experiments have shown that in urine the assay sensitivity, as defined by the limit of detection, is reduced compared to assay buffer, due to the fact that various proteins, salts and pH fluctuations in human urine impair assay sensitivity. In an alternative approach, different concentrators were applied based on ion exchange and size exclusion in order to prevent the dilution of the sample. Analytical urine was spiked with marker candidates treated with concentrators, based on the principle of size exclusion (millipore represents components >10kDa) and ion exchange (proteospin), and then the assay was performed on ARChip Epoxy as usual. Concentration based on size exclusion (device Millipore) does improve the assay sensitivity for selected markers.
Efforts for further improving assay sensitivity of the selected panel of biomarkers included the use of aptamers instead of capture antibodies.
In another approach to further enhance assay sensitivity we evaluated carbon nanoparticles and upconverted nanophosphor (UCNP) as labels in biomarker microarrays. UCNPs are reported to provide several advantages over conventional fluorophors which may suffer from serious drawbacks including autofluorescence, photobleaching, spectral overlap, and low photostability especially in biological fluids despite their high sensitivity and wide dynamic range. Also, the use of multiple excitation sources for different fluorophores renders the detection instrumentation more complex. Photon upconverting nanophosphors (UCNPs) are inorganic lanthanide-doped nanocrystals capable of converting lowerenergy near-infrared (NIR) excitation into higher-energy visible light emission. Because of the unique optical process producing anti-Stokes photoluminescence, the UCNPs can be measured free of autofluorescence and scattered excitation light, enabling highly sensitive assays. Additionally, they are completely photostable.
While the implementation of carbon nanoparticles did not further improve the detection limits of biomarkers in urine, the use of UCNPs allowed for up to 10times enhanced detection limits for some of the biomarkers when working on white substrates such as nitrocellulose.
Cell analytics
An assay platform combining cancer cell recognition and counting in a novel fluorescence urine cytology with a cell based bioassay of biomarkers was developed. For recognition the cell nuclei were stained with a blue life cell nuclear stain. The cytology platform contains an automated fluorescence imaging device using a three colour detection channel. The blue channel is used for morphological evaluation of the nuclei and green and red channels are used for surface biomarker evaluation. All evaluations were performed using Merkofix fixed cytospin slides. This technology was used to achieve save transport of the slides.
For quantification of surface biomarker detection the intensities are normalized to the nuclei staining. For the different biomarkers the relative intensities were calibrated relative to each other as this is also important for the determination of optimum antibody concentrations. All the readout and evaluation of the data are done by a novel computer program integrated in the device. Characterization of cells by morphology and expression of surface biomarkers are both integrated into the evaluation program. This procedure allowed sampling of several probes for every patient and gave good results in the evaluation of nuclei morphology and surface biomarkers. All probes where analysed against classical PAP cytology as gold standard. The assays for all biomarkers and considered primary/secondary antibody systems were developed and tested. One slide of a series was first evaluated by classical Papanicolaou staining. From the number of cells observed on the slide the necessary amount of primary antibodies was estimated. The next slides were than stained using an appropriate mixture of blue HOECHST nuclear stain and the respective primary antibodies. After washing the slide was treated with the solution of the secondary antibodies. After washing the images were taken in a standard fluorescence microscope or in the Onkocell-C equipment.
Image of a fluorescence array using immobilized capture antibodies/aptamers. Multiplexed assay characteristics for 4 selected biomarkers. Cluster of polymorph tumor cells stained with CK20 antibody (green). The nuclei are shown in blue.
1.3.5 Experimental results
Multiplexed platform
We in the discovery study evaluated 13 molecular biomarkers and 10 clinical phenotype parameters with a study design specifically focusing on BC recurrence diagnosis. Using generalized linear regression models as statistics procedure for deriving classification functions we concluded on 6 molecular biomarkers and 3 clinical parameters.
Relevant to note is that only combined parameter use in a classification function provided reasonable AUC of 0.9 however, in leave-one-out cross validation dropping to 0.70 clearly indicating the need of an independent validation study.
Testing our classifier in an independent sample cohort provided as area under the curve AUC = 0.86 (in leave-one-out cross validation seeing an AUC of 0.76).
From given analysis a combination of clinical parameters together with molecular parameters is needed for identifying clinically meaningful AUC values. Of further relevance is the need of normalizing biomarker readout the urinary creatinine. The present classifier includes as parameters from the clinical side the number of past recurrences, status of BCG therapy, and staging at first diagnosis; molecular biomarkers include Ecadh, IL8, ErbB2, IL6, EN2, VEGF.
Biomarker discovery pipeline
Further discovery efforts identified additional biomarker candidates, including EN2, S100A12 and HMGB1.
EN2 case(BC):control(healthy) S100A12, different stages of BC Tissue staining for HMBG1 expression
From given data on additional biomarker candidates a further extension of the DIPROMON classifier is to be considered.
1.3.6 Integrative platform
We derived the following blueprint of the DIPROMON integrated platform, each element being covered by a SOP:
The blueprint aims at covering BC diagnosis as well as disease progression, together with a predictive component for covering drug mechanism of action interference.
According to given classifiers both, clinical as well as molecular phenotyping is to be captured, seeing definition of clinical base data together with the SOP for sample procurement (plasma, urine). Biomarker quantitation sees different implementations covering secreted proteins in urine (point of care testing; multiplexed ELISA platform) together with cell-based analytics (urine; cell count/morphology as well as surface marker staining). A software component is realized taking the different clinical and molecular parameters as input for given classifiers, and also allowing computation of novel classifiers according to adopted input parameters.
AXO multiplexed biomarker quantitation chip holding DIPROMON markers LIODetect®UBC with spiked urine samples Onkocell-II imaging system.
AXO DIPROMON marker panel, multiplexed
The biomarker multiplex assay using integrated system developed is based on a sandwich assay format at the bottom of a 96-well plate. Spots are produced using a piezoelectric nano-dispenser through non-contact printing. Typical size of spot is 150-200µm in diameter. The bottom plate multiplex assay is then processed using any liquid handling robotic platform equipped with a robotic arm and a liquid handling (from 4 to 96 channels). All steps of the assay protocol are then automated. The reading of the assay results is performed using a 96-well plate imager (for example the CLAIR system from Sensovation) and the image is captured in gray-scale. The analysis of the assay images is performed using the AXOWare program which gives results either as gray-scale numbers or marker concentration values.
LIODetect®UBC as a Diagnostic test: using EN2 as bladder cancer marker
The LIODetect®UBC Test under development is under oncology section point-of-care and laboratory tests, and shall an aid in the diagnosis of bladder cancer. The LIODetect®UBC Competitive Test is a non-invasive assay, performed on a single urine sample that detects elevated levels of EN2 protein. The test can be performed in a physician’s office with results delivered during the patient visit, allowing a rapid, accurate and cost-effective means of aiding the detection of bladder cancer in patients at risk, when used in conjunction with standard diagnostic procedures. The EN2 test is a novel, rapid competitive assay. The competition will take place between the sample EN2 and the coated EN2 to binding to the gold-labelled specific EN2 antibodies. If low EN2 or absence in the sample the high binding ratio of the labelled anti-EN2 to the coated EN2 on solid phase and that mean more density color and versus-wise if high level of EN2 in the sample.
Next, we focus on evaluation trials in different cohorts to determine a universal threshold. Our aim is to scale up manufacture of inexpensive and rapid-test device and this is expected to happen early 2017 and after that we will be in good position to set POC-specific exploitation plan.
Onkocell-II imaging system
The system includes a fluorescence or absorption reader of biomarker array assays. Therefore, cytology evaluation can be directly combined with biomarker assays in one instrument.
Software integration
Software components include classifier generation/application according to input from the experimental platforms, and molecular process models, assigned molecular biomarkers and drug mechanism of action interference.
Software module for linking biomarker readout and clinical parameters in a classification function for BC recurrence risk. MoA interference with BC pathology aspects: doxorubicin and biomarkers included in the classification function. MoA interference with BC pathology aspects: progesterone and biomarkers included in the classification function.
With project completion the DIPROMON team implemented a diagnosis/prognosis/prediction blueprint for bladder cancer.
Procedures are in place for further expanding in classification of disease, and tools/SOPs are ready for executing further validation studies.
Potential Impact:
Potential impact
1.4.1 Contributions to State-of-the-Art
With cystoscopy as gold standard in BC diagnosis/monitoring DIPROMON established alternative procedures aimed at adding accuracy and clinical decision making. Important to note is the clear need of further validation studies for determining robustness and clinical utility of established procedures.
On scientific grounds DIPROMON allowed establishing assays, test procedures and application data covered in scientific publications. On top we added to an understanding of BC molecular pathology, with superficial disease as well as muscle invasive disease characteristics being published.
1.4.2 Socio-economic impact
Materials, assays and patent applications on the ground of DIPROMON foreground allow involved SMEs adding opportunities to their commercial portfolio. Specifically to note is implementation of a prototype point-of-care device as well as a multiplexed ELISA array for BC diagnosis, and a software component for optimized disease pathology:drug mechanism of action interference screening now embedded in a scientific software platform. With this given status the project directly supported SMEs of the DIPROMON consortium.
The broader perspective of precision medicine is considered as path forward in clinical management of disease, also identified as a major R&D strategy in Europe. Impact is first of all patient-centric via avoiding unnecessary procedures and medication, with the example of a non-invasive test for recurrence diagnosis combined with assessment of drug response. Clinical implementation will further see reduction of direct costs combined with socio-economic impact of finally limiting level of morbidity.
1.4.3 Scientific impact
“Precision medicine” has become a buzzword in the medical domain, i.e. holding decision support in hand for implementing optimal therapy measures. However, practical implementation sees limits, mainly due to lack of causality in disease onset/progression, expression on the level of clinical phenotype parameters, and involvement of biomarkers.
The ideal scenario solely uses molecular parameters in such matching procedure as direct proxy of disease pathology and drug mechanism of action. DIPROMON classification functions on disease progression (resembling the clinical finding of recurrence) still combine clinical phenotype parameters and molecular proxies. Still with given work we added to a general stratification concept, exemplified on bladder cancer.
List of Websites:
http://www.surrey.ac.uk/dipromon/(si apre in una nuova finestra)
Background:
Bladder cancer (BC) is a prevalent malignancy imposing a major burden on patients and healthcare. While first line procedures are successful in reducing tumor load, BC is characterized by a high probability for recurrence, eventually developing into muscle invasive disease. In consequence tight clinical monitoring is implemented with cystoscopy as gold standard.
Challenge:
As with most clinical diagnosis procedures sensitivity as well as specificity see limits also in BC, hence improving BC management via adding supportive, ideally non-invasive procedures sees clear clinical need. DIPROMON was set out to improve diagnostic accuracy for recurrence and disease progression via identifying a panel of molecular biomarkers. Yet the biggest challenge is identification of appropriate measures triggered by such diagnosis – i.e. biomarker guided precision in drug use.
Integrative concept:
For adding to precision medicine exemplified on BC we followed a concept of modelling disease pathology and drug mechanism of action as molecular process models, and selected biomarker candidates at model interference. Including experimental testing on disease progression-associated molecular processes promises identification of biomarker candidates with on top of diagnosis/prognosis also support in drug selection as downstream consequence.
Technology implementation:
Along DIPROMON procedures we first established BC pathology models including annotation regarding association with disease progression, from there picking a panel of biomarker candidates. Assembling experimental materials further triggered assay development including multiplexed ELISAs and cell-based analytics (morphology and surface marker stains), embedded in a software framework for computing classification functions on risk for recurrence/disease progression.
Results:
On the basis of an established DIPROMON biobank we evaluated biomarker-based classification functions for BC recurrence diagnosis, a top-ranked classifier forwarded to an independent validation study. AUC values obtained indicate capturing a relevant portion of BC recurrence events. Involved biomarkers – due to their assignment to defined molecular processes – allow analyzing interference to given therapy regimes (as BCG), and further allowed us screening alternative drugs promising beneficial interference with BC pathology.
IP and dissemination:
All relevant findings of DIPROMON were published in peer-reviewed journals and presented at scientific conferences for assuring provision of our results to the scientific domain. Specific aspects were filed for IP protection.
Outlook:
A blueprint of the integrated DIPROMON analysis platform including SOPs for all procedures is in place and ready for further validation studies.
Project Context and Objectives:
1.2 Description of project context and objectives
1.2.1 Bladder cancer diagnosis and therapy – clinical perspective
Urinary bladder cancer (BC) is a common disease worldwide. About 2.7 million people have a history of BC. The highest incidence of BC is in developed countries, especially in North America and Europe. In USA alone, about 570,000 people suffered from bladder cancer in 2011, with steadily increasing prevalence. The WHO estimates that there are 330,000 new cases annually worldwide. In Germany, about 26,000 new cases of bladder cancer are diagnosed every year.
The most common clinical presentation is blood in the urine/hematuria, which is the principle initial symptom in 90% of bladder cancer patients. Usually this is painless and the blood may be visible to the naked eye (gross hematuria) or can be seen only under the microscope (microscopic hematuria). Frequently, the diagnosis of bladder cancer is delayed because bleeding is intermittent or attributed to other causes, such as urinary tract infection or blood thinners. However, a substantial proportion of these patients will have a significant problem such as kidney stones or tumors, urinary tract obstruction and bladder cancer.
Bladder cancer has high survival rates but also high recurrence rates, which leads to a requirement for an active surveillance regime for people with a history of bladder cancer. Even after full excision of superficial bladder tumors, there is up to a 75% chance of recurrence. Although most recurrences are also superficial, bladder cancer is a multi-focal disease and new tumors may not necessarily occur at the same location. Accordingly, a post-surgery monitoring program is always initiated. Guidelines from the American Urology Association recommend follow up is undertaken quarterly, reducing to annual checks at five years from first diagnosis. In most cases surveillance tests are performed by cystoscopy, a routine but unpleasant procedure for the patient.
Because of long-term survival and the need for lifelong routine monitoring and treatment, the cost per patient of bladder cancer from diagnosis to death is ranging from $96,000-187,000 in the USA. Overall, bladder cancer is the fifth most expensive cancer in terms of total medical care expenditures, accounting for almost 3.7 billion $ in direct costs just in the USA.
Current laboratory diagnosis of bladder cancer is done by examination of the urine, vagina, or rectum. The following tests and procedures are mostly used:
• Urine analysis: A test to check urine content including sugars, proteins, red/white blood cells
• Urine cytology: Microscopic examination of urine to check for abnormal cells
• Cystoscopy: A procedure to look inside the bladder and urethra to check for abnormal areas
Despite its low sensitivity (35% to 40%) in the detection of urothelial carcinoma of all grades and stages (17% in grade 1, 61% in grade 2 and 90% in grade 3), urine cytology remains the most commonly used non-invasive test. If the urinary cytology is positive, then transitional cell cancer of the urothelium is almost certainly present. However, cytologic examinations may be negative in up to half of patients with bladder cancer; thus, a negative study does not rule out bladder cancer.
In this context the objectives of DIPROMON were to add options aside invasive cystoscopy in surveillance by non-invasive cytology in combination with biomarker tests. The set of biomarkers is aimed to give information on the probability of recurrence and invasiveness of the tumor and prognosis on the effectiveness of therapy. Especially the individual response to BCG immunotherapy in respect to chemotherapy should be assessed.
1.2.2 Project objectives
With project filing we defined the following specific objectives to be tackled with DIPROMON:
• Design of a novel patient stratification concept
• Building tools and procedures for allowing implementing such stratification concept
• Validating the concept in the clinical context of bladder cancer
Patient stratification concept
The general topic being followed in DIPROMON experienced significant momentum in the translational research community during project runtime, coining “stratified” as well as “precision” medicine. The general notion is to improve molecular phenotyping on top of clinical phenotyping, i.e. adding or ideally replacing clinical descriptors of disease presentation with molecular mechanistic parameters. Reasoning is apparent: a clinical presentation exhibits a molecular pathophysiology, itself being a composite of molecular processes. A molecular biomarker in its definition serves as proxy for the status of molecular processes, hence assembling a panel of biomarkers reflecting pathologic processes of a clinical presentation serve for detailed molecular phenotyping. Such strategy holds a major benefit in the context of analyzing drug treatment and response, being the pivotal aim for improving diagnosis and prognosis (making only sense if triggering a clinical consequence). In essence, a drug modulates a target’s activity, triggering changes in downstream molecular processes. Hence both, disease pathology as well as drug mechanism of action can be described on the level of molecular processes, with molecular biomarkers as anchors. On this basis a development strategy towards realizing stratified/personalized medicine holds the following elements:
• Biomarker-based molecular phenotyping
• Clinical phenotyping (holding established factors for describing disease presentation and progression)
• Molecular models of drug mechanism of action
Combining these elements (with improved diagnosis and prognosis as inherent element) promises improved precision in drug use and molecular mechanistic guidance in drug development.
DIPROMON tools and procedures
In-silico models of pathology and drug MoA
Rich molecular data space is available for characterizing bladder cancer pathology, however, needing consolidation on the level of molecular processes for allowing rational selection of a corresponding biomarker panel. This needs an algorithmic procedure for traversing molecular feature sets into molecular model representations. Same applied on feature sets characterizing drug mechanism of action allows molecular process interference analysis for linking drug effect and disease pathology with molecular biomarkers.
Biomarker materials
Candidate biomarkers and respective antibodies need production and purification for implementing assays.
Multiplexed assays
With focus on urine as sample matrix assays have to be established, according to the biomarker panel concept needing further integration in multiplexed formats. For prototyping and subsequent clinical use automated readout technology needs to be in place.
Cell-based analytics
Both, morphological characteristics (to be analyzed via specific staining) as well as molecular characteristics (expression of relevant receptors) are to complement biomarker-based profiling of samples.
Integration software
Software is to be established taking a clinical/molecular data set as input for computing a score according to a classification function.
SOPs
For assuring standardized application of components and workflow set of SOPs is to be established
Validation, bladder cancer
For validating the concept a biobank of bladder cancer samples was to be established, including detailed clinical phenotyping at baseline (first diagnosis), medication regimes and follow-up data on results from surveillance according to clinical practise.
Establishing and validating tools and procedures was conducted in a two-step fashion, a discovery study for defining relevant clinical and molecular parameters and respective classification functions, followed by a validation study.
Primary focus in-line with clinical need was recurrence diagnosis, for reflecting our concept combined with treatment regimes further coupled with prognosis towards development of muscle invasive disease.
Project Results:
Description of main S&T results/foreground
1.3.1 Project workflow and specific goals
With focus on bladder cancer DIPROMON was organized in six RTD work packages complemented by management as well as dissemination/exploitation.
Central procedure implemented biomarker candidate selection, on such basis establishing experimental tools for marker quantitation, entering a discovery study for biomarker shortlisting and classifier generation, further forwarded to a validation study.
In parallel the conceptual fundament of our stratification approach was elaborated and continuous biomarker profiling for optimizing diagnosis/prognosis/prediction was realized.
The DIPROMON biobank provided sample grounds for biomarker and classifier testing, altogether meeting in technology prototype development and integration blueprint together with respective standard operating procedures.
With this structure we implemented the central aims of DIPROMON (design of a novel patient stratification concept; building tools and procedures allowing implementation; validation) in project runtime.
1.3.2 Stratification concept
With precision medicine and personalization considered as key strategy for improving efficacy of approved drugs as well as improving the development path in drug development, patient phenotyping on a combined clinical and molecular level in regard to the phenotype in focus is proposed, with a critical role of early stage clinical testing.
Disease pathology and drug effect resemble molecular process perturbance and respective interference with drug mechanism of action, demanding description on the level of molecular phenotyping. In preclinical models and early stage trials in human a link of molecular processes, clinical phenotype parameters and outcome via molecular biomarkers bi is required. Specifically early stage clinical trials serve for evaluation of a drug’s molecular mechanistic impact in human via including molecular process biomarkers already in study design. On such basis adaptive trial designs as biomarker-based enrichment can be pursued from phase II trials onwards, linking molecular processes to drug effect via use of predictive biomarkers.
Here, diagnosis and prognosis are to be considered intrinsically different to prediction of drug response. While expanded phenotyping may provide improved accuracy in detecting onset and progression of disease, prediction claims at adding precision in drug selection and response with respect to beneficial outcome on a personalized level. Clinical together with molecular phenotyping, in particular via utilizing molecular biomarkers, serve as most relevant tools in this context. However, diagnosis as well as prognosis is not sufficient in case we aim at controlling development of an endpoint of a molecular system S, i.e. some P(S), in the given scenario implying interference with S via utilizing a specific drug and its mechanism of action. For assuring control we need specific knowledge about the structure of S, otherwise we lack means for rational interference. Prediction rests on diagnosis of an aggregate property of a molecular system S at a certain time t (clinical follow-up) and integrates prognosis with structure (in a molecular process/pathway context). Purpose of prediction is to prognose the consequence of interference with S via a selected drug.
Consider a system S reflecting a dynamical hierarchy. Implicit question is what molecular characteristics xi to modify – including the consequences this may have on the context of xi, being the entire dynamical hierarchy, for allowing such optimization. The drug in the first place introduces an object constraint on the specific target(s) it is binding, in further consequence propagating this constraint in the dynamical hierarchy. Due to the relation constraints, consequence of addressing the target may become apparent on various levels of the hierarchy, eventually also including a change in a clinical parameter. In case the drug is applied on the level of S – although seeing molecular phenotypes Si, Sj (an appreciated fact specifically in oncology) – a further consequence may be variance in drug response, eventually detected on the level of a clinical parameter, but per design detected on the level of molecular biomarkers out of X, resembling the requirements of precision medicine.
Let us consider some S holding a drug target and two molecular biomarkers. Addressing the target via a drug (the drug imposing an object constraint on the first order object) may trigger a change in some higher level observable O2 (e.g. a molecular pathway status) further reflecting in some clinical phenotype parameter O3. In a different composition of S no consequence is seen on the level of O2, but still exhibiting an equivalent readout in O3. Such scenario reflects two molecular phenotypes Si, Sj out of S, in case S is defined on the level of O3. If defined on the level of O2 two strata of S can be determined, namely Si and Sj.
O2 in this example setting resembles the status of a molecular pathway, with biomarkers xi and xj characterizing two distinct molecular processes being constituents of the pathway. According to the given scheme the abundance (e.g. with respect to a certain threshold) of both biomarkers are indicative for O2 in the one situation, whereas just one molecular process is identified as active in the other situation (e.g. with a down-regulated xj), in consequence seeing difference in O2. Although O3 is equivalent in both cases, altering O2 may be the essential parameter regarding a clinical endpoint P(S). A practical molecular example is the apoptosis pathway with promoting (determined by xi) and inhibiting (determined by xj) molecular process components.
Aiming at interfering with a drug addressing xt, hence altering the process seen as relevant via the proxy xi, will in the example case see consequence on P(S) in Si, but no effect for subjects embedded in Sj. In consequence, we expect variance in drug response on the level of S, a fact which can be determined when including xi and xj in stratification for response when targeting xt with a certain drug.
According to this concept any S needs to be captured on the level of molecular processes. Equivalently, drug mechanism of action needs to be assessed on a molecular process level for allowing a fit-for-purpose evaluation utilizing molecular biomarkers as process proxy.
For leveraging on this concept bladder cancer together with drug mechanism of action need to be modelled on a molecular process level.
1.3.3 Bladder cancer pathology and biomarker candidates
In line with the stratification concept we consolidated a repository of molecular profiles, signatures and individual features characterizing BC, split into superficial as well as muscle invasive disease. We in total extracted molecular features from 5 SNP studies, 3 mRNA transcriptomics studies, 1 miRNA transcriptomics study, 2 proteomics studies, and 3 metabolomics studies resulting in a set of 897 relevant proteins after mapping miRNA to their targets and metabolites to enzymes. Additionally we executed literature mining (NCBI PubMed) for retrieving individual molecular features reported as relevant in BC, finally concluding on a set of about 1,300 molecular features linked with BC.
For traversing the molecular feature set into a molecular process model we mapped the molecular feature set on a hybrid protein interaction network, followed by using topological criteria for extracting molecular processes and their biological relations. Molecular features remaining in the molecular model were screened for annotation in the biomarker context (split into diagnostic/prognostic), combined with Omics profile analysis on disease progression according to statistical testing. With such biomarker annotation in hand we analyzed the molecular pathway content of each process for defining molecular processes of specific relevance in disease progression, followed by selection of representative biomarker candidates for each molecular process deemed being involved in progressive disease.
According to this procedure we concluded on a first set of biomarker candidates, complemented by a consortium efforts in identifying further candidates deemed relevant but not covered in the process model.
Further analytics focused on cell surface markers including proteins of the human epidermal growth factor receptor (HER) family, Uroplakins and CK20/CD45. The last pair is used to discriminate cancer cells from white blood cells. On top, CK20 appears also to be a prognostic marker on bladder cancer cells.
We from this analytics steps concluded on a set of about 30 biomarker candidates to be further followed materials development for ELISA and cell staining.
For complementing the deductive analysis on the molecular model basis also an explorative strategy was followed via transcriptomics profiling of samples (Affymetrix® Human Gene 1.1 ST Array Strip using Affymetrix GeneAltas System).
1.3.4 Materials and technologies
Biobanking
According to clinical needs the following key areas of interest were defined:
• Adding precision/changing procedures for BC recurrence diagnosis
• Estimating risk for disease progression, muscle invasive disease
• Estimating response to medications
Primary focus of assay/product development was on diagnosing recurrence for complementing regular surveillance procedures. Biobanking of samples combined with deep annotation of clinical parameters primarily focused on this scenario, further complemented by sampling advanced stages. Compliance with ethics requirements were explicitly handled and reported in the context of project deliverables. The full DIPROMON biobank finally included in total 249 transitional cell bladder cancer and 80 muscle invasive bladder patient samples We have also collected urine and blood samples from 114 healthy age matched volunteers as controls.
new diagnosis Recurrence Neg Neg previous pos HVC G1 pTa G1 pT1 G2 pTa G2 pT1 G3 pTa G3 pT1 TBC
urine 66 79 43 61 114 6 2 90 24 6 26 15
Blood Paxgene 7 15 1 22 112 2 0 13 1 0 1 0
Plasma 7 15 1 22 112 2 0 13 1 0 1 0
Table of samples of urine and blood from Non-muscle-invasive bladder cancer patients or control (grade and stage at sample).
G2 pT2 G3 pT2a G3 pT2b G3 pT3a G3 pT3b G3 T4a G3 T4b
6 36 0 18 1 17 2
Table of samples from invasive bladder cancer patients (grade and stage at sample)
Materials
For the spectrum of biomarker candidates a pragmatic procedure was implemented with respect to material availability for designing assays, in part including commercially available reagents. Still, significant work was involved in generating new materials including recombinant production of proteins as well as corresponding antibodies. By this a set of reagents became available being of general interest in BC research.
EN2 Engrailed 2 UPK3A Uroplakin-3a FGFR3 Fibroblast growth factor receptor 3 CD44 molecule (Indian blood group)
IL8 Interleukin 8 MMP9 Matrix metallopeptidase 9 VEGFA Vascular endothelial growth factor A
VIM Vimentin
EGFR Epidermal growth factor receptor APOE Apolipoprotein E Cyfra 21-1Cytokeratin Fragment-19 MYC Myelocytomatosis viral oncogene homolog (MYC)
PTGS2 Prostaglandin-endoperoxide synthase 2 EN2_full Length with 2 his-tags Engrailed 2 FGFR3-First N-ter domain Truncated FGFR3 FGFR3-Second C-ter domain Truncated FGFR3
IL-6 Interleukin 6 UPK1B Uroplakin 1B KRT20 keratin 20 CDH1 E-cadherin
Multiplexed ELISA
Multiplexed sandwich immunoassays on the ARChip Epoxy platform were established to quantitatively measure protein biomarkers. Analyte binding was detected with biotinylated antibodies in combination with Streptavidin/Dy647. Assay conditions were optimized with respect to the chip platform (ARChip Epoxy), printing buffer (PBS at pH 7.2 supplemented with Na-deoxycholate), probe concentration (capture antibodies spotted on the chip), assay buffer (commercial available LowCross buffer) and signal output.
Recovery experiments have shown that in urine the assay sensitivity, as defined by the limit of detection, is reduced compared to assay buffer, due to the fact that various proteins, salts and pH fluctuations in human urine impair assay sensitivity. In an alternative approach, different concentrators were applied based on ion exchange and size exclusion in order to prevent the dilution of the sample. Analytical urine was spiked with marker candidates treated with concentrators, based on the principle of size exclusion (millipore represents components >10kDa) and ion exchange (proteospin), and then the assay was performed on ARChip Epoxy as usual. Concentration based on size exclusion (device Millipore) does improve the assay sensitivity for selected markers.
Efforts for further improving assay sensitivity of the selected panel of biomarkers included the use of aptamers instead of capture antibodies.
In another approach to further enhance assay sensitivity we evaluated carbon nanoparticles and upconverted nanophosphor (UCNP) as labels in biomarker microarrays. UCNPs are reported to provide several advantages over conventional fluorophors which may suffer from serious drawbacks including autofluorescence, photobleaching, spectral overlap, and low photostability especially in biological fluids despite their high sensitivity and wide dynamic range. Also, the use of multiple excitation sources for different fluorophores renders the detection instrumentation more complex. Photon upconverting nanophosphors (UCNPs) are inorganic lanthanide-doped nanocrystals capable of converting lowerenergy near-infrared (NIR) excitation into higher-energy visible light emission. Because of the unique optical process producing anti-Stokes photoluminescence, the UCNPs can be measured free of autofluorescence and scattered excitation light, enabling highly sensitive assays. Additionally, they are completely photostable.
While the implementation of carbon nanoparticles did not further improve the detection limits of biomarkers in urine, the use of UCNPs allowed for up to 10times enhanced detection limits for some of the biomarkers when working on white substrates such as nitrocellulose.
Cell analytics
An assay platform combining cancer cell recognition and counting in a novel fluorescence urine cytology with a cell based bioassay of biomarkers was developed. For recognition the cell nuclei were stained with a blue life cell nuclear stain. The cytology platform contains an automated fluorescence imaging device using a three colour detection channel. The blue channel is used for morphological evaluation of the nuclei and green and red channels are used for surface biomarker evaluation. All evaluations were performed using Merkofix fixed cytospin slides. This technology was used to achieve save transport of the slides.
For quantification of surface biomarker detection the intensities are normalized to the nuclei staining. For the different biomarkers the relative intensities were calibrated relative to each other as this is also important for the determination of optimum antibody concentrations. All the readout and evaluation of the data are done by a novel computer program integrated in the device. Characterization of cells by morphology and expression of surface biomarkers are both integrated into the evaluation program. This procedure allowed sampling of several probes for every patient and gave good results in the evaluation of nuclei morphology and surface biomarkers. All probes where analysed against classical PAP cytology as gold standard. The assays for all biomarkers and considered primary/secondary antibody systems were developed and tested. One slide of a series was first evaluated by classical Papanicolaou staining. From the number of cells observed on the slide the necessary amount of primary antibodies was estimated. The next slides were than stained using an appropriate mixture of blue HOECHST nuclear stain and the respective primary antibodies. After washing the slide was treated with the solution of the secondary antibodies. After washing the images were taken in a standard fluorescence microscope or in the Onkocell-C equipment.
Image of a fluorescence array using immobilized capture antibodies/aptamers. Multiplexed assay characteristics for 4 selected biomarkers. Cluster of polymorph tumor cells stained with CK20 antibody (green). The nuclei are shown in blue.
1.3.5 Experimental results
Multiplexed platform
We in the discovery study evaluated 13 molecular biomarkers and 10 clinical phenotype parameters with a study design specifically focusing on BC recurrence diagnosis. Using generalized linear regression models as statistics procedure for deriving classification functions we concluded on 6 molecular biomarkers and 3 clinical parameters.
Relevant to note is that only combined parameter use in a classification function provided reasonable AUC of 0.9 however, in leave-one-out cross validation dropping to 0.70 clearly indicating the need of an independent validation study.
Testing our classifier in an independent sample cohort provided as area under the curve AUC = 0.86 (in leave-one-out cross validation seeing an AUC of 0.76).
From given analysis a combination of clinical parameters together with molecular parameters is needed for identifying clinically meaningful AUC values. Of further relevance is the need of normalizing biomarker readout the urinary creatinine. The present classifier includes as parameters from the clinical side the number of past recurrences, status of BCG therapy, and staging at first diagnosis; molecular biomarkers include Ecadh, IL8, ErbB2, IL6, EN2, VEGF.
Biomarker discovery pipeline
Further discovery efforts identified additional biomarker candidates, including EN2, S100A12 and HMGB1.
EN2 case(BC):control(healthy) S100A12, different stages of BC Tissue staining for HMBG1 expression
From given data on additional biomarker candidates a further extension of the DIPROMON classifier is to be considered.
1.3.6 Integrative platform
We derived the following blueprint of the DIPROMON integrated platform, each element being covered by a SOP:
The blueprint aims at covering BC diagnosis as well as disease progression, together with a predictive component for covering drug mechanism of action interference.
According to given classifiers both, clinical as well as molecular phenotyping is to be captured, seeing definition of clinical base data together with the SOP for sample procurement (plasma, urine). Biomarker quantitation sees different implementations covering secreted proteins in urine (point of care testing; multiplexed ELISA platform) together with cell-based analytics (urine; cell count/morphology as well as surface marker staining). A software component is realized taking the different clinical and molecular parameters as input for given classifiers, and also allowing computation of novel classifiers according to adopted input parameters.
AXO multiplexed biomarker quantitation chip holding DIPROMON markers LIODetect®UBC with spiked urine samples Onkocell-II imaging system.
AXO DIPROMON marker panel, multiplexed
The biomarker multiplex assay using integrated system developed is based on a sandwich assay format at the bottom of a 96-well plate. Spots are produced using a piezoelectric nano-dispenser through non-contact printing. Typical size of spot is 150-200µm in diameter. The bottom plate multiplex assay is then processed using any liquid handling robotic platform equipped with a robotic arm and a liquid handling (from 4 to 96 channels). All steps of the assay protocol are then automated. The reading of the assay results is performed using a 96-well plate imager (for example the CLAIR system from Sensovation) and the image is captured in gray-scale. The analysis of the assay images is performed using the AXOWare program which gives results either as gray-scale numbers or marker concentration values.
LIODetect®UBC as a Diagnostic test: using EN2 as bladder cancer marker
The LIODetect®UBC Test under development is under oncology section point-of-care and laboratory tests, and shall an aid in the diagnosis of bladder cancer. The LIODetect®UBC Competitive Test is a non-invasive assay, performed on a single urine sample that detects elevated levels of EN2 protein. The test can be performed in a physician’s office with results delivered during the patient visit, allowing a rapid, accurate and cost-effective means of aiding the detection of bladder cancer in patients at risk, when used in conjunction with standard diagnostic procedures. The EN2 test is a novel, rapid competitive assay. The competition will take place between the sample EN2 and the coated EN2 to binding to the gold-labelled specific EN2 antibodies. If low EN2 or absence in the sample the high binding ratio of the labelled anti-EN2 to the coated EN2 on solid phase and that mean more density color and versus-wise if high level of EN2 in the sample.
Next, we focus on evaluation trials in different cohorts to determine a universal threshold. Our aim is to scale up manufacture of inexpensive and rapid-test device and this is expected to happen early 2017 and after that we will be in good position to set POC-specific exploitation plan.
Onkocell-II imaging system
The system includes a fluorescence or absorption reader of biomarker array assays. Therefore, cytology evaluation can be directly combined with biomarker assays in one instrument.
Software integration
Software components include classifier generation/application according to input from the experimental platforms, and molecular process models, assigned molecular biomarkers and drug mechanism of action interference.
Software module for linking biomarker readout and clinical parameters in a classification function for BC recurrence risk. MoA interference with BC pathology aspects: doxorubicin and biomarkers included in the classification function. MoA interference with BC pathology aspects: progesterone and biomarkers included in the classification function.
With project completion the DIPROMON team implemented a diagnosis/prognosis/prediction blueprint for bladder cancer.
Procedures are in place for further expanding in classification of disease, and tools/SOPs are ready for executing further validation studies.
Potential Impact:
Potential impact
1.4.1 Contributions to State-of-the-Art
With cystoscopy as gold standard in BC diagnosis/monitoring DIPROMON established alternative procedures aimed at adding accuracy and clinical decision making. Important to note is the clear need of further validation studies for determining robustness and clinical utility of established procedures.
On scientific grounds DIPROMON allowed establishing assays, test procedures and application data covered in scientific publications. On top we added to an understanding of BC molecular pathology, with superficial disease as well as muscle invasive disease characteristics being published.
1.4.2 Socio-economic impact
Materials, assays and patent applications on the ground of DIPROMON foreground allow involved SMEs adding opportunities to their commercial portfolio. Specifically to note is implementation of a prototype point-of-care device as well as a multiplexed ELISA array for BC diagnosis, and a software component for optimized disease pathology:drug mechanism of action interference screening now embedded in a scientific software platform. With this given status the project directly supported SMEs of the DIPROMON consortium.
The broader perspective of precision medicine is considered as path forward in clinical management of disease, also identified as a major R&D strategy in Europe. Impact is first of all patient-centric via avoiding unnecessary procedures and medication, with the example of a non-invasive test for recurrence diagnosis combined with assessment of drug response. Clinical implementation will further see reduction of direct costs combined with socio-economic impact of finally limiting level of morbidity.
1.4.3 Scientific impact
“Precision medicine” has become a buzzword in the medical domain, i.e. holding decision support in hand for implementing optimal therapy measures. However, practical implementation sees limits, mainly due to lack of causality in disease onset/progression, expression on the level of clinical phenotype parameters, and involvement of biomarkers.
The ideal scenario solely uses molecular parameters in such matching procedure as direct proxy of disease pathology and drug mechanism of action. DIPROMON classification functions on disease progression (resembling the clinical finding of recurrence) still combine clinical phenotype parameters and molecular proxies. Still with given work we added to a general stratification concept, exemplified on bladder cancer.
List of Websites:
http://www.surrey.ac.uk/dipromon/(si apre in una nuova finestra)